Professional Documents
Culture Documents
Constantin Milici
Gheorghe Drăgănescu
J. Tenreiro Machado
Introduction
to Fractional
Differential
Equations
www.pdfgrip.com
Volume 25
Series editor
Albert C. J. Luo
Southern Illinois University
Edwardsville, IL, USA
www.pdfgrip.com
Introduction to Fractional
Differential Equations
123
www.pdfgrip.com
J. Tenreiro Machado
Department of Electrical Engineering
Polytechnic of Porto
Porto, Portugal
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
www.pdfgrip.com
Preface
v
www.pdfgrip.com
vi Preface
Preface vii
References
1. Atanackovič, T. M., Pilipovič, S., Stankovič, B., & Zorica, D. (2014). Frac-
tional calculus with applications in mechanics: Wave propagation, impact and
variational principles. London: Wiley-ISTE.
2. Atanackovič, T. M., Pilipovič, S., Stankovič, B., & Zorica, D. (2014). Frac-
tional calculus with applications in mechanics: Vibrations and diffusion pro-
cesses. London: Wiley-ISTE.
3. Barabási, A.-L., & Stanley, H. E. (2002). Fractal concepts in surface growth
(2nd ed.). Cambridge: Cambridge University Press.
4. Biswas, K., Bohannan, G., Caponetto, R., Mendes Lopes, A., & Tenreiro
Machado, J. A. (2017). Fractional-order devices. SpringerBriefs in nonlinear
circuits. New York: Springer.
5. Bueno-Orovio, A., Kay, D., Grau, V., Rodriguez, B., & Burrage, K. (2014).
Fractional diffusion models of cardiac electrical propagation: Role of structural
heterogeneity in dispersion of repolarization. Journal of the Royal Society
Interface, 11, 20140352.
6. Das, S. (2011). Functional fractional calculus. Berlin: Springer.
7. de Oliveira, E. C., & Tenreiro Machado, J. A. (2014). A review of definitions
for fractional derivatives and integral. Mathematical Problems in Engineering,
2014, Article ID 238459.
8. Deü, J.-F., & Matignon, D. (2010). Simulation of fractionally damped mechan-
ical systems by means of a Newmark-diffusive scheme. Computers & Mathe-
matics with Applications, 59, 1745–1753.
9. Diethelm, K., Ford, N. J., & Freed, A. D. (2002). Simulation of fractionally
damped mechanical systems by means of a Newmark-diffusive scheme. Non-
linear Dynamics, 59, 3–22.
www.pdfgrip.com
viii Preface
Preface ix
27. Ortigueira, M. D., Tenreiro Machado, J., & Trujillo, J. J. (2017). Fractional
derivatives and periodic functions. International Journal of Dynamics and
Control, 5, 72–78.
28. Oustaloup, A. (2014). Diversity and non-integer differentiation for system
dynamics. Hoboken: Wiley-ISTE.
29. Petráš, I. (2011). Fractional-order nonlinear systems modeling, analysis and
simulation. Beijing/Heidelberg: Higher Education Press/Springer.
30. Podlubny, I. (1998). Fractional differential equations: An introduction to frac-
tional derivatives, fractional differential equations, to methods of their solution
and some of their applications. Mathematics in science and engineering. San
Diego: Academic Press.
31. Post, E. L. (1930). Generalized differentiation. Transactions of the American
Mathematical Society, 32, 723–781.
32. Povstenko, Y. (2015). Linear fractional diffusion-wave equation for scientists
and engineers. Cham: Birkhäuser.
33. Povstenko, Y. (2015). Fractional thermoelasticity. Cham: Springer.
34. Pritchard, R. H., & Terentjev, E. M. (2017). Oscillations and damping in the
fractional Maxwell materials. Journal of Rheology, 61, 187–203.
35. Ray, S. S. (2016). Fractional calculus with applications for nuclear reactor
dynamics. Boca Raton: CRC Press.
36. Rossikhin, Yu. A., & Shitikova, M. V. (2004). Analysis of the viscoelastic
rod dynamics via models involving fractional derivatives or operators of two
different orders. The Shock and Vibration Digest, 36, 3–26.
37. Sabatier, J., Agrawal, O. P., & Tenreiro Machado, J. A. (Eds.). (2007). Advances
in fractional calculus theoretical developments and applications in physics and
engineering. Dordrecht: Springer.
38. Samko, S. G., Kilbas, A. A., & Marichev, O. I. (Eds.). (1993). Fractional
integrals and derivatives: Theory and applications. Yverdon: Gordon and
Breach Science Publishers.
39. Saxena, R. K., Saxena, R., & Kalla, S. L. (2010). Solution of spacetime
fractional Schrödinger equation occurring in quantum mechanics. Fractional
Calculus & Applied Analysis, 13, 177–190.
40. Sheng, H., Chen, Y. & Qiu, T. (Eds.). (2012). Fractional processes and
fractional-order signal processing: Techniques and applications. London:
Springer.
41. Shingareva, I. K., & Lizárraga-Celaya, C. (Eds.). (2011). Solving nonlinear
partial differential equations with maple and mathematica. Wien: Springer.
42. Singh, R., Davies, P., & Bajaj, A. K. (2003). Identification of nonlinear and
viscoelastic properties of flexible polyurethane foam. Nonlinear Dynamics, 34,
319–346.
43. Tenreiro Machado, J. (2001). Discrete-time fractional-order controllers. Frac-
tional Calculus & Applied Analysis, 4, 47–66.
44. Tenreiro Machado, J. A. (2010). Entropy analysis of integer and fractional
dynamical systems. Nonlinear Dynamics, 62, 371–378.
www.pdfgrip.com
x Preface
45. Tenreiro Machado, J. A., Mainardi, F., & Kiryakova, V. (2010). A poster about
the old history of fractional calculus. Fractional Calculus & Applied Analysis,
13, 447–454.
46. Tenreiro Machado, J., Kiryakova, V., & Mainardi, F. (2011). Recent history
of fractional calculus. Communications in Nonlinear Science and Numerical
Simulation, 16, 1140–1153.
47. Tenreiro Machado, J., Lopes, A. M., Duarte, F. B., Ortigueira, M. D., & Rato,
R. T. (2014). Rapsody in fractional. Fractional Calculus & Applied Analysis,
17, 1188–1214.
48. Tenreiro Machado, J. A., Mainardi, F., Kiryakova, V., & Atanacković, T. (2016).
Fractional calculus: D’où venons-nous? Que sommes-nous? Où allons-nous?
Fractional Calculus & Applied Analysis, 17, 1074–1104.
49. Valério, D., & Costa, J. S. (2013). An introduction to fractional control.
London: CRC Press.
50. Valério, D., Tenreiro Machado, J. A., & Kiryakova, V. (2014). Some pioneers of
the applications of fractional calculus. Fractional Calculus & Applied Analysis,
17, 552–578.
51. Weilbeer, M. (2005). Efficient numerical methods for fractional differential
equations and their analytical background. PhD thesis, Facultät für Mathematik
und Informatik, Technischen Univerisität Braunschweig, Braunschweig.
52. Wheeler, N. (1997). Construction and physical application of fractional calcu-
lus. Technical report, Reed College Physics Department.
53. Yang, X.-J., Baleanu, D., & Tenreiro Machado, J. A. (2013). Mathematical
aspects of the Heisenberg uncertainty principle within local fractional Fourier
analysis. Boundary Value Problems, 131, 1–16.
54. Yang, X.-J., Baleanu, D., Srivastava, H. M., & Tenreiro Machado, J. A.
(2013). On local fractional continuous wavelet transform. Abstract and Applied
Analysis, 2013, 5.
55. Zhang, X., Wei, C., Liu, Y., & Luo, M. (2014). Fractional corresponding oper-
ator in quantum mechanics and applications: A uniform fractional Schrödinger
equation in form and fractional quantization methods. Annals of Physics, 350,
124–136.
56. Zhou, Y., Ionescu, C., & Tenreiro Machado, J. A. (2015). Fractional dynamics
and its applications. Nonlinear Dynamics, 80, 1661–1664.
www.pdfgrip.com
Contents
1 Special Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Euler’s Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Gamma Function. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.2 Beta Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2 Integral Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3 Mittag-Leffler Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Function E(t, α, a). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2 Fractional Derivative and Fractional Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.1 Fractional Integral and Derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3 The Laplace Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1 Calculus of the Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2 Calculus of the Original Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2.1 Calculus of Original Using Residues . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2.2 Calculus of Original with Post’s Inversion Formula . . . . . . . . . . 36
3.3 The Properties of the Laplace Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.1 The Property of Linearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.2 Similarity Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.3 The Differentiation and Integration Theorems . . . . . . . . . . . . . . . . 37
3.3.4 Delay Theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3.5 Displacement Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3.6 Multiplication Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3.7 Properties of the Inverse Laplace Transform . . . . . . . . . . . . . . . . . . 39
3.4 Laplace Transform of the Fractional Integrals and Derivatives . . . . . . . 43
3.4.1 Fractional Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.4.2 Fractional Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
xi
www.pdfgrip.com
xii Contents
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
www.pdfgrip.com
Acronyms
FC Fractional calculus
FD Fractional derivative
FDE Fractional differential equation
FI Fractional integral
LT Laplace transform
RK Runge–Kutta
RK2 Second order Runge–Kutta
RK4 Fourth order Runge–Kutta
VIM Variational iteration method
List of Symbols
E(t, α, a) E function
Jp (t) Bessel function of first kind
B Beta function
erfc Complementary error function
C Set of complex numbers
det Determinant
erf Error function
γ Euler’s constant
Ei Exponential integral function
Gamma function
erfi Imaginary error function
L Laplace transform
max Maximum
min Minimum
Eα One parameter Mittag-Leffler function
N Set of natural numbers
R Set of real numbers
Re Real part
Res Residues
sup Supremum
Eα,β Two parameter Mittag-Leffler function
www.pdfgrip.com
Chapter 1
Special Functions
In this chapter several special functions used in the follow-up of the book are
presented briefly. More details about these functions can be found in [1, 3, 4, 8].
1
where I1 = e−x x p−1 dx is convergent.
0
1 L. Euler (1707–1783).
2 1 Special Functions
4
y
–4 –3 –2 –1 0 1 2 3 4
x
–2
–4
–6
x p−1
1 ≤ x ⇒ x p−1 e−x ≤ e−x/2 ⇔ x p−1 ≤ ex/2 ⇔ ≤ 1.
ex/2
x p−1
Because lim = 0, we have:
x→∞ ex/2
∞ ∞
e−x x p−1 dx ≤ e−x/2 dx = 2e−1/2 .
1 1
∞ ∞
−x p
Proof (p+1) = e x dx = − e−x x p ]∞
0 +p e−x x p−1 dx = p(p).
0 0
3. The following relations are also valid:
(1) = 1,
(n + 1) = n!,
(0) = +∞.
4. For p = −n it results:
(−n + 1)
(−n) =
−n
(−n + 2) (−n + 3) (0)
= = = . . . = (−1)n = (−1)n ∞.
n(n − 1) n(n − 1)(n − 2) n!
(p + 1)
5. Taking account that the function can be written as (p) = , it
p
results that the function can be defined also for negative values of p, in the
interval −1 < p < 0.
If −n < p < −(n − 1), then from (1.3) it results:
(p + n)
(p) = .
p(p + 1) . . . (p + n − 1)
(−1)n (α)
(α − n) = .
(1 − α)(2 − α) . . . (n − α)
4 1 Special Functions
or:
1 (2m − 1)! 1 (2m)! 1
m+ = = .
2 2m 2 m!22m 2
8. The following particular values for function can be useful for calculation
purposes:
1 √
= π,
2
1 √
− = −2 π ,
2
3 1 1 1 1√
= 1+ = = π,
2 2 2 2 2
4! 1
5 1 2 3√
= 2+ = 4
= π,
2 2 2! 2 4
1 1 1 4 . . . (3m − 2) 1
= 2.678938 m+ = ,
3 3 3m 3
2 2 2 5 . . . (3m − 1) 2
= 1.354118 m+ = ,
3 3 3m 3
1 1 1 5 . . . (4m − 3) 1
= 3.625600 m+ = ,
4 4 4m 4
3 3 3 7 . . . (4m − 1) 3
= 1.225417 m+ = .
4 4 4m 4
9.
(p + 1) p
= ,
(q + 1)(p − q + 1) q
www.pdfgrip.com
Then, we obtain:
∞
−x p−1
k x k p−1
(p) = e x dx = lim 1− x dx.
0 k→∞ 0 k
But:
(k + 1)p
lim = 1.
k→∞ kp
From:
1 1 2p 3p . . . (k + 1)p
(p) = lim ,
p k→∞ (1 + p)(1 + p/2) . . . (1 + p/k) 1p 2p . . . k p
it follows:
∞
1 p p −1
∞
1
1 (k + 1)p 1
(p) = = 1 + 1+ ,
p (1 + p/k) k p p k k
k=1 k=1
6 1 Special Functions
valid, excepting the values Re(p) = 0, −1, −2, . . . , −n, . . . , where Real
part (Re) represents the real part and n ∈ N.
Here the symbol Euler’s constant (γ ) represents the Euler’s constant, given by:
p
1
γ = lim − ln p = 0.577215663 . . .
p→∞ k
k=1
and finally:
1 1 p p
p
= lim = epγ lim e−p (1 + p)e− 2 . . . e− k 1 +
(p) k→∞ k (t) k→∞ k
∞
p −p
= peγp 1+ e k.
k
k=1
(1 − p)
But (−p) = . It results:
p
∞
1 1 p2
=p 1− 2 .
(p) (1 − p) k
k=1
∞
p2
Because sin(nπ ) = π n 1 − 2 it results the reflection property.
k
k=1
Another Proof
∞ ∞
−t p−1
(p)(1 − p) = e t dt e−s s 1−p−1 ds
0 0
∞ ∞ p
−(t+s) t
= e t −1 dtds.
0 0 s
1
Figure 1.2 depicts (x) .
8 1 Special Functions
y
1
–4 –3 –2 –1 0 1 2 3
x
–1
–2
In the following we will enumerate the basic properties of the Beta function:
1. For every p > 0 and q > 0, we have:
2. For every p > 0 and q > 1, the Beta function B satisfies the property:
q −1
B(p, q) = B(p, q − 1).
p+q −1
Proof
1
B(p, q) = x p−1 (1 − x)q−1 dx,
0
1 xp x p (1 − x)q−1 1 q − 1 1 p
B(p, q) = (1 − x) d =
q−1
+ x (1 − x)q−2 dx
0 p p 0 p 0
q − 1 1 p−1 q − 1 1 p−1
= x (1 − x) dx −
q−2
x (1 − x)q−1 dx
p 0 p 0
q −1 q −1
= B(p, q − 1) − B(p, q).
p p
www.pdfgrip.com
(p)(q)
B(p, q) = .
(p + q)
resulting:
D[t, s]
dtds = dxdy = xdxdy,
D[x, y]
1 1
(p)(q) = e−x (xy)p−1 x q−1 (1 − y)q−1 xdxdy
0 0
∞ 1
−x p+q−1
= e x dx y p−1 (1 − y)q−1 dy,
0 0
4. For every p > 0, and for the natural number n, it can be proved
1 · 2 · 3 . . . (n − 1)
B(p, n) = B(n, p) = ,
p(p + 1) . . . (p + n)
and also:
1
B(p, 1) = .
p
www.pdfgrip.com
10 1 Special Functions
(n − 1)!(m − 1)!
B(m, n) = .
(m + n − 1)!
1+x
and replacing u = , we have:
2
p−1 p−1
2 (p) 1 1 1+x 1−x 1 1 p−1
= = 1 − x2 dx,
(2p) 2 −1 2 2 22p−1 −1
resulting
1
22p−1 2 (p) = 2(2p) (1 − x 2 )p−1 dx.
0
1 1 1 √
But B ,p = 2 (1 − x )
2 p−1
dx, = π , resulting finally:
2 0 2
1
(p)
1 2
22p−1 2 (p) = (2p)B , p = (2p) .
2 1
+p
2
6. Triplication formula:
1
33p− 2 1 2
(3p) = √ (p) p + p+ .
π 2 3
12 1 Special Functions
∞
zk
Eα (z) = Re(α) > 0.
(αk + 1)
k=0
∞
∞
zk
zk
E1,1 (z) = = = ez , (1.5)
(k + 1) k!
k=0 k=0
∞
zk∞ ∞
zk 1 zk+1 ez − 1
E1,2 (z) = = = = , (1.6)
(k + 2) (k + 1)! z (k + 1)! z
k=0 k=0 k=0
∞
∞ ∞
zk zk 1 zk+2 ez − 1 − z
E1,3 (z) = = = 2 = ,
(k + 3) (k + 2)! z (k + 2)! z2
k=0 k=0 k=0
(1.7)
∞
zk
E1,0 (z) = = ze ,
z
(1.8)
(k)
k=0
eat √
E1, 3 (at) = √ erf( at), (1.9)
2 at
1 √ √
E1, 1 (at) = √ + eat at erf( at), (1.10)
2 π
1 eat √ 2
E1, 5 (at) = √ erf( at) − √ , (1.11)
2 at at π
1 1 √ √
E1,− 1 (at) = − √ + (at) √ + at eat erf( at) , (1.12)
2 2 π π
1
Eα,β (t) = + tEα,α+β (t), (1.13)
(β)
www.pdfgrip.com
∞
∞
∞
tk t k+1 tk
Eα,β (t) = = =
(αk + β) (α(k + 1) + β) (αk + α + β)
k=0 k=−1 k=−1
1
= + tEα,α+β (t),
(β)
(k + m)! ∞
(m) dm tk
Eα,β (z) = Eα,β (z) = . (1.14)
dzm k! (αk + αm + β)
k=0
We prove schematically the equivalence between the forms (1.15) and (1.16). We
denote the integral:
t
1
I = τ α−1 ea(t−τ ) dτ
(α) 0
∞
1 t a k (t − τ )k
= τ α−1
dτ
(α) 0 k!
k=0
∞
1 ak t k t τ k
= τ α−1
1− dτ .
(α) k! 0 t
k=0
τ
For u = we obtain:
t
1
t τ k
I1 = τ α−1 1 − dτ = t α−1 uα−1 (1 − u)k du
0 t 0
www.pdfgrip.com
14 1 Special Functions
1 t α (α)(k + 1)
= tα uα−1 (1 − u)k du = t α B(α, k + 1) = .
0 (α + k + 1)
It results:
∞
(at)k
I = tα = t α E1,α+1 (at) = E(t, α, a).
(k + α + 1)
k=0
In (1.16) we take α = 1
2 and u2 = aτ . It results after calculations:
eat √
E1, 3 (at) = √ erf(a t).
2 at
Using the definition we obtain:
∞
(at)k
E1, 1 (at) = .
2
k=0 k + 12
1 1 √ √
= √ + atE1, 3 (at) = √ + ateat erf( at)
π 2 π
References
1. Abramowitz, M., & Stegun, I. A. (Eds.). (1965). Handbook of mathematical functions. Dover
books on mathematics. New York: Dover Publications.
2. Brychkov, Y. A. (2008). Handbook of special functions derivatives, integrals, series and other
formulas. Boca Raton: Chapman and Hall/CRC.
3. Erdélyi, A. (1954). Tables of integral transforms (Vols. 1–2). New York: McGraw-Hill Book
Company.
4. Erdélyi, A. (Ed.). (1955). Higher transcendental functions (Vols. I–III). New York: McGraw-
Hill Book Company
www.pdfgrip.com
References 15
5. Gorenflo, R., Kilbas, A. A., Mainardi, F., & Rogosin, S. V. (2014). Mittag-Leffler functions,
related topics and applications. Springer monographs in mathematics. Heidelberg: Springer.
6. Lorenzo, C. F., & Hartley, T. T. (2017). The fractional trigonometry: With applications to
fractional differential equations and science. Hoboken: Wiley & Sons, Inc.
7. Nikiforov, A. F., & Ouvarov, V. (1976). Eléments de la théorie des fonctions spéciales. Moscow:
Mir Publishers.
8. Podlubny, I. (1998). Fractional differential equations: An introduction to fractional derivatives,
fractional differential equations, to methods of their solution and some of their applications.
Mathematics in science and engineering. San Diego: Academic Press.
www.pdfgrip.com
Chapter 2
Fractional Derivative and Fractional
Integral
Definition (Fractional Integral of Order α) For every α > 0 and a local inte-
grable function f (t), the right FI of order α is defined:
t
1
α
a It f (t) = (t − u)α−1 f (u)du, −∞ ≤ a < t < ∞. (2.1)
(α) a
For particular values of the a and b parameters, the following cases are known:
• Riemann1 : a = 0, b = +∞
• Liouville2 : a = −∞, b = 0.
Definition (Fractional Derivative of Order α) For every α, and n = α
the
Riemann–Liouville derivative of order α can be defined as:
1 d n t
a Dt f (t) = (t − u)n−α−1 f (u) du.
α
(2.3)
(n − α) dx a
b b
f (x)a Dxα g(x)dx = g(x)x Dbα f (x) dx. (2.5)
a a
1 φ(x)ψ(t)
f (x, t) = ,
(α) (x − t)1−α
n (t − a)α−k
a It [a Dt f (t)] = f (t) −
α−k
(b) α α D
a t f (t) ,
t=a (α − k + 1)
k=1
(t − a)α−1
(c) D[a Itα f (t)] = a Itα [Df (t)] + f (a),
(α)
α+p
D k f (a)(t − a)α+k
p−1
(d) a Itα f (t) = a It [D p f (t)] + , where p is a positive
(α + k + 1)
k=0
integer.
D k f (a)(t − a)α+k
p−1
(e) D p [a Itα f (t)] = a Itα [D p f (t)] + , where p is positive
(α + k + 1)
k=0
integer.
Proof
(a) Integrating by parts, it results:
t t
1 1
a It
α+1
[Df (t)] = (t − u)α f (u)du = [(t − u)α f (u)]
(α + 1) a (α + 1) a
t
α
+ (t − u)α−1 f (u) du,
(α + 1) a
(t − a)α α t
a It
α+1
[Df (t)] = − f (a) + (t − u)α−1 f (u) du.
(α + 1) α(α) a
and we make the change of variable: u = t − x 1/α (see also [3]). We obtain:
(t−a)α
1
α
a It f (t) = f (t − x 1/α ) dx.
(1 + α) a
Hence:
(t − a)α−1 f (a)
D[a Itα f (t)] = + a Itα [Df (t)].
(α)
(t − a)α f (a)
a It
α+1
[D 2 f (t)] = a Itα+1 [Df (t)] − .
(α + 1)
Df (a) (t − a)α−2
D 2 [a Itα f (t)] = D[a Itα f (t)] + .
(α − 1)
[Df (a) (t − a)α−2 ] [Df (a) (t − a)α−1 ]
D 2 [a Itα f (t)] = a Itα f (t) + + ,
(α − 1) (α)
β α+β
α
a It a It f (t) = a It f (t).
we obtain:
t t
1 1
I= (t − u)α−1 (u − v)β−1 f (v) dudv,
(α) (β) a v
u = v + z(t − v),
but:
1 (α)(β)
(1 − z)α−1 zβ−1 du = B(α − 1, β − 1) = .
0 (α + β)
Finally, it results:
t
1 α+β
I= (t − v)α+β−1 f (v)dv = a It f (t).
(α + β) a
Theorem 4
β α−β
(a) a Dtα a It f (t) = a Dt f (t).
m
(t − a)α−k
β α−β β−k
(b) a It a Dt f (t) = a It f (t) −
α
a Dt f (t) .
(α + 1 − k) t=a
k=1
where m = β
+ 1.
www.pdfgrip.com
m (t − a)−α−k
β α+β β−k
(c) α
a Dt D
a t f (t) = D
a t f (t) − D
a t f (t) .
t=a (1 − α − k)
k=1
Proof
(a)
d n n−α β d n n−(α−β)
β α−β
α
a Dt a It f (t) = a It a It f (t) = a It = a Dt .
dt n dt n
(b)
t
β 1 β
I = a Itα a Dt f (t) = (t − u)α−1 a Dt f (u) du
(α) a
t
1 d β
I= (t − u)α a Dt f (u) du
(α + 1) a dt
t
(t − a)α β 1 d
β−1
I= − D
a t f (t) + (t − u)α+1 a Dt f (u) du
(α+1) t=a (α+2) a dt
...
where: C1 and C2 are constants and f (t) and g(t) are two arbitrary functions.
www.pdfgrip.com
Proof
t
1
a Dt [C1 f (t) + C2 g(t)] = (t − y)α−1 [C1 f (y) + C2 g(y)] dy
α
(α) a
t
1
= C1 (t − y)α−1 f (y)dy
(α) a
t
1
+C2 (t − y)α−1 g(y)dy
(α) a
= C1 a Itα f (t) + C2 a Itα g(t).
Theorem 6 If the function f (t) possess continuous derivative, then for α >
0, n = α
+ 1:
n−1 (k) t
f (t − a)k−α 1
α
a It f (t) = + (t − y)n−α−1 f (n) (y) dy.
(k + 1 − α) (n − α) a
k=0
p
(−1)k p!t p−k
u = (t − (t − u)) =
p p
(t − u)k ,
k!(p − k)!
k=0
p t
p! 1 1
I= (−1)k t p−k (t − u)α+k−1 f (u) du,
(p − k)! (α) 0 k!
k=0
p
(α + k − 1) . . . α d k p α+k
I= (−1)k t I f (t),
k! dt k
k=0
p α d k
I= (−1)k t p I α+k f (t).
k dt k
k=0
d n n−α p
(b) I = D α [t p f (t)] = I [t f (t)]. We obtain:
dt n
p
−n+α n d k+j
n
I= t p D α−k−j f (t)
k j dt k=j
k=0 j =0
i
−n+α n
p
d i p n−i
I= i
t D f (t) ,
dt i−j j
i=0 j =0
but
i
−n+α n α
= ,
i−j j i
j =0
thus:
p k
α d
D α [t p f (t)] = t p D α−k f (t).
k dt k
k=0
5 M. Caputo (1967–).
www.pdfgrip.com
n−1
f (k) (a)
a Dt f (t) = a Dt f (t) + (t − a)k−α .
α C α
(1 + k − α)
k=0
n−1
tk
f (t) = f (k) (0) + Rn−1 ,
(k + 1)
k=0
where:
t f (n) (y)(t − y)n−1 1 t
Rn−1 = dy = f (n) (y)(t − y)n−1 dy = I n f (n) .
0 (n − 1)! (n) 0
n−1
t k−α
D α f (t) = f (k) (0) + I n−α f n (t),
(k − α + 1)
k=0
n−1
t k−α
D α f (t) = f (k) (0) + D α f (t).
(k − α + 1)
k=0
where α
= n.
Observation
If C is a constant, then:
C α
0 Dt C = 0,
6 B. Taylor (1685–1731).
www.pdfgrip.com
C x −α
α
0 Dt C = , α = 1, 2, . . . .
(1 − α)
In what follows we note the Caputo derivative in the origin, simply, using the
notation D α f (x).
Theorem 9 If n − 1 < α < n, where n ∈ N, and α ∈ R, then:
It results:
1 (t − y)n−α t
D α f (t) = −f (n) (y)
(n − α) n−α 0
t
1
+ (t − y) n−α (n+1)
f (y)dy .
n−α 0
it results:
t
1
D α f (t) = f (n) (0) + f (n+1) (y)(t − y)n−α dy ,
(n − α + 1) 0
www.pdfgrip.com
t t
lim D f (t) = f
α (n)
(0) + f (n+1) (y)dy ] = f (n) (0) + f (n) (y) = f (n) (t),
α→n 0 0
t
lim D α f (t) = f (n) (0) + f (n+1) (y)(t − y)dy
α→n−1 0
t
= f (n) (0)t + (t − y)f (n) (y) = f (n−1) (t) − f (n−1) (0).
0
Example 1 Let us calculate the FD for α > 0, n − 1 < α < n, β > n − 1 of the
function f (t) = t β using the definitions, for the case:
1. Riemann–Liouville.
2. Caputo in the origin, using the definition.
Solution
1. For the Riemann–Liouville derivative, we can write:
t
1 dn
I =D t = α β
uβ (t − u)n−α−1 du,
(n − α) dt n 0
and we take:
u = v t, du = t dv.
It follows:
t
1 dn
I = (vt)β [(1 − v)t]n−α−1 t dv
(n − α) dt n 0
t
1 dn
= (1 − v)n−α−1 v β t n−α+β dv,
(n − α) dt n 0
t
1 d n n−α+β
I= (1 − v)n−α−1 v β t dv,
(n − α) 0 dt n
but
dn λ (λ + 1) λ−n
t = t ,
dt n (λ − n + 1)
1
B(p, q) = v p−1 (1 − v)q−1 dv,
0
www.pdfgrip.com
so that it results:
1 (n − α + β + 1) −α+β 1
I= t (1 − v)n−α−1 v β dv,
(n − α) (−α + β + 1) 0
1 (n − α)(β + 1)
(1 − v)n−α−1 v β dv = B(n − α, β + 1) = ,
0 (n − α + β + 1)
(β + 1)
Dα t β = I = t β−α .
(−α + β + 1)
du = t dv,
(β + 1) 1
I= (uv)β−n [(t − v)n−α−1 ]t dv.
(n − α)(β − n + 1) 0
Finally, we obtain:
f (t) = (t − a)β .
u−a
= v,
t −a
du = (t − a)dv,
www.pdfgrip.com
allows to calculate:
(t − a)α+β 1 (t − a)α+β
I= (1 − v)α−1 v β dv = B(α, β + 1),
(α) 0 (α)
(β + 1)
I= (t − a)α+β .
(α + β + 1)
(β + 1) dn (β + 1)
Df = (t − a)β+n−α = (t − a)β−α .
(β + n − α + 1) dt n (β − α + 1)
Theorem 10 If n − 1 < α < n and if f (t) satisfy the conditions of the Taylor
theorem, then:
∞
f (k) (0)
D f (t) =
α
t k−α .
(k − α + 1)
k=0
Proof Because f (t) satisfy the conditions of the Taylor theorem, we can apply the
Taylor expansion:
∞ ∞
f (k) (0) k f (k) (0) k
f (t) = t = t .
k! (k + 1)
k=0 k=0
The FD will be
∞
∞
f (k) (0) α k f (k) (0) (k + 1) k−α
D α f (t) = D t = t ,
(k + 1) (k + 1) (k − α + 1)
k=0 k=0
and finally:
∞
f (k) (0)
D α f (t) = t k−α .
(k − α + 1)
k=0
and taking
r
(−1)k r! r−k
τ = [t − (t − τ )] =
r r
t (t − r)k ,
k!(r − k)!
k=0
we obtain:
r t
(−1)k r! r−k 1 1
I α [t r f (t)] = t (t − τ )α+k−1 f (τ )dτ.
k!(r − k)! (α) 0 k!
k=0
But:
t
1
I α+k
f (t) = (t − τ )α+k−1 f (τ )dτ,
(α + k) 0
Finally, we obtain:
r
(α + k − 1) . . . α d k r α+k
I α [t r f (t)] = (−1)k t I f (t)
k! dt k
k=0
r k
−α d
= t r I α+k f (t).
k dt k
k=0
Observation
For 0 < α < 1, f (0) = 0, we have:
t
1
I α f (t) = (t − u)α−1 f (u) du,
(α) 0
t t
α 1
I α f (t) = (t − u)α−1 f (u) du = f (u) (du)α .
α(α) 0 (α + 1) 0
www.pdfgrip.com
References 31
References
1. Caputo, M. (1999). Lessons on seismology and rheological tectonics. Technical report, Univer-
sitá degli Studi La Sapienza, Rome.
2. Ishteva, M., Scherer, R., & Boyadjiev, L. (2003). On the Caputo operator of fractional calculus
and C-Laguerre functions. Technical report, Bulgarian Ministry of Education and Science, Grant
MM 1305/2003.
3. Miller, K. S., & Ross, B. (1993). An introduction to the fractional calculus and fractional
differential equations. New York: John Wiley & Sons.
4. Valério, D., & Costa, J. S. (2013). An introduction to fractional control. London: CRC Press.
5. Podlubny, I. (1998). Fractional differential equations: An introduction to fractional derivatives,
fractional differential equations, to methods of their solution and some of their applications.
Mathematics in science and engineering. San Diego: Academic Press.
www.pdfgrip.com
Chapter 3
The Laplace Transform
is called the Laplace1 integral, or Laplace transform (LT), or image of the original
function f (t). In the follow-up we denote by L[f (t)] = F (s) or simply by Laplace
transform (L) the Laplace transform. In Table 3.1 the LT of some elementary usual
functions are listed.
The corresponding inverse Laplace transform is [1, 4]:
γ +it
1
f (t) = lim F (s)est dt = L−1 [F (s)], (3.2)
2π i t→∞ γ −it
√
where i = −1 and γ ∈ R, so that the contour path of integration is contained in
the convergence region.
1 bt 1 −bt
Example 3 Find the image of f (t) = te + te .
2 2
1 1 s 2 + b2
F (s) = + = .
2(s − b)2 2(s + b)2 (s 2 − b2 )2
www.pdfgrip.com
If we denote by L[f (t)] = F (s), then if we consider all residues of the function
F (s)est , denoted by r1 , r2 , . . . rn we can use the theorem:
f (t) = r1 + r2 + . . . + rn .
The residues, denoted by Residues (Res), can be calculated using the following
procedure (theorem):
If a is a simple pole of the function, then:
1
Res[est F (s)] = lim [(s − a)n est F (s)](n−1) .
a (n − 1)! s→a
1
F (s) = .
(s − 3)2 (s + 1)
est e−t
r1 = lim (s + 1) = ,
t→−1 (s − 3)2 (s + 1) 16
st
d e
r2 = lim
t→3 ds s+1
te3t − e3t
= ,
16
resulting:
s2
F (s) = .
s4 −1
www.pdfgrip.com
Solution The function F has singularities: 1, −1, −i, i. The residues will be:
s 2 est et
r1 = Res F (s)est = lim (s − 1) = ,
1 s→1 (s − 1)(s + 1)(s 2 + 1) 4
s 2 est e−t
r2 = Res F (s)est = lim (s + 1) = − ,
−1 s→−1 (s − 1)(s + 1)(s 2 + 1) 4
s 2 est −e−it
r3 = Res F (s)est = lim (s + i) = ,
−i s→−i (s 2 − 1)(s − i)(s + i) 4i
s 2 est eit
r4 = Res F (s)est = lim (s + i) = .
i s→i (s 2 − 1)(s − i)(s + i) 4i
It results finally
1 et − e−t 1 eit − e−it
f (t) = + ,
2 2 2 2i
or:
1
f (t) = (sinh t + sin t).
2
n!
F (s) = n!s −1−n .
s n+1
2 E. Post (1897–1954).
www.pdfgrip.com
k!
lim √ k k e−k = 1,
k→∞ 2π k
it results:
n −n n n k n n
f (t) = t e lim 1+ 1+ 1+ = t n.
k→∞ k k k
In this section, we will use the notations F (s) = L[f (t)] and L(s) = G[g(t)]. In
the follow-up are discussed properties of the LT.
3 J. Stirling (1692–1770).
www.pdfgrip.com
F (s)
L[g] = .
s
and, finally:
d
F (n) (s) = [F (n−1) (s)].
ds
The convolution product of two functions f (t) and g(t) is designated by the symbol
∗. We have:
t
(f ∗ g)(t) = f (τ )g(t − τ ) dτ,
0
−1 s −(α−1)
L = t α−1 E1,α (at) = E(t, α − 1, a), (3.7)
s−a
−1 s −α
L = tE(t, α, a) − αE(t, α + 1, a), (3.8)
(s − a)2
−1 s −α 1 α(α + 1)
L = t 2 E(t, α, a) − αtE(t, α + 1, a) + E(t, α + 2, a),
(s − a) 3 2 2
(3.9)
n+k
∞ (−a) k
1
L−1 = t α(n+1)−1 k
t k(α−β) ,
(s α + as β )n+1 [k(α − β) + (n + 1)α]
k=0
(3.10)
where 0 < β ≤ α.
n+k
∞
∞ (−b)n (−a)k
sγ
L−1 = t α−γ −1 k
t k(α−β)+nα ,
s α + as β + b [k(α − β) + (n + 1)α − γ ]
n=0 k=0
(3.11)
1 a k
∞
∞
ak (kβ + α) s −(α−β)
= = = .
(kβ + α) s kβ+α sα sβ sβ − a
k=0 k=0
d
L[t E(t, α, a) − αE(t, α + 1, a)] = − L[E(t, α, a)] − αL[E(t, α + 1, a)]
ds
−α
d s s −(α+1) 1
=− −α = α .
ds s − a s−a s (s − a)2
www.pdfgrip.com
1 d2 d α(α + 1)
= 2
L [E(t, α, a)] + α L [E(t, α + 1, a)] + L [E(t, α + 2, a)]
2 ds ds 2
−α
1 d2 s d s −(α+1) α(α + 1) s −(α+2) 1
= +α + = α .
2 ds 2 s−a ds s−a 2 s−a s (s − a)3
For the identities (3.10) and (3.11) the reader can use the reference [5].
Proof of the identity (3.10). We will apply the well-known identity [4]:
1 n+k
∞
= (−x)k .
(1 + x) n k
k=0
It follows:
1 1 1
= α n+1
(s α + as β )n+1 (s ) a n+1
1 + αβ
s
1 n+k −a k
∞
= .
(s α )n+1 k s α−β
k=0
sγ sγ 1
=
(s + s a)
α β n+1 s α(n+1) a n+1
1 + α−β
s
1 n+k −a k
∞ ∞
n+k (−a)k
= α(n+1)−γ = ,
s k s α−β k s α(n+1)+k(α−β)−γ
k=0 k=0
sγ
∞
n+k
∞
(−a)k
= (−b) n
,
s α + as β + b k s α(n+1)+k(α−β)−γ
n=0 k=0
www.pdfgrip.com
n+k
∞
∞ (−b)n (−a)k
sγ
L−1 = t α−γ −1 k
t k(α−β)+nα .
s α + as β + b [k(α − β) + (n + 1)α − γ ]
n=0 k=0
Solution
and finally:
Solution
Y (s) 1 1
Y (s) = + ⇒ Y (s) = .
s s s−1
Solution
1 1 s 1 1
Y (s) = − 2 = , ⇒ Y (s) = .
s2 + 1 s s +1 s(s 2 + 1) s
Solution
1 1 1
Y (s) = Y (s) − ; ⇒ Y (s) = .
s−1 s−1 s−2
If α > 0, the Riemann–Liouville and Caputo FI are the same for both cases:
t
1
I = I f (t) =
α
(t − y)α−1 f (y)dy.
(α) 0
1 F (s)
L[I ] = L[t α−1 ]L[f (t)] = α .
(α) s
– The Riemann–Liouville FD is
n t
1 dt
L Dtα f (t) = L (t − u) n−α−1
f (u)du
(n − α) d n t 0
n
dt
=L I n−α f (t) ,
d nt
n−1
L[Dtα f (t)] = s α F (s) − s n−α−1 [D k I n−α f (t)]t=0 .
k=0
and
n−1
L[D α f (t)] = s α F (s) − s n−α−1 f (k) (0).
k=0
Exercise 5 For the function f (t) = t 2 , calculate the Caputo L[D α ]. It results:
1
1. α = ,
2
1
2. α = − .
2
Solution
1.
1
L[D 1/2 t 2 ] = L[2t],
1− 21
s
2
L[D 1/2 t 2 ] = 5
,
s2
3 3
2 2t 2 8t 2
D 1/2 t 2 = L−1 5
= = √ .
s2 52 3 π
www.pdfgrip.com
References 45
we define:
∞
Lα [f (t)] = Fα (s) = Eα (−s α t α )f (t) (dt)α .
0
we have:
Lα (f (t) ∗ g(t))α = Lα [f (t)] Lα [g(t)].
References
1. Abramowitz, M., & Stegun, I. A. (Eds.). (1965). Handbook of mathematical functions. Dover
books on mathematics. New York: Dover Publications.
2. Caputo, M. (1999). Lessons on seismology and rheological tectonics. Technical report,
Universitá degli Studi La Sapienza, Rome.
3. Caputo, M., & Fabrizio, M. (2015). Damage and fatigue described by a fractional derivative
model. Journal of Computational Physics, 293, 400–408.
4. Debnath, L., & Bhatta, D. (2007). Integral transforms and their applications. Boca Raton:
Chapman and Hall/CRC.
5. Erdélyi, A. (Ed.). (1955). Higher transcendental functions (Vols. I–III). New York: McGraw-
Hill Book Company
6. Jumarie, G. (2009). Laplaces transform of fractional order via the Mittag-Leffler function and
modified Riemann-Liouville derivative. Applied Mathematics Letters, 22, 1659–1664.
7. Post, E. L. (1930). Generalized differentiation. Transactions of the American Mathematical
Society, 32, 723–781.
www.pdfgrip.com
8. Poularikas, A. D. (Ed.). (2000). The transforms and applications handbook (2nd ed.). CRC
Press: Boca Raton.
9. Schiff, J. L. (Ed.). (1999). The laplace transform, theory and applications. New York: Springer.
10. Wheeler, N. (1997). Construction and physical application of fractional calculus. Technical
report, Reed College Physics Department.
www.pdfgrip.com
Chapter 4
Fractional Differential Equations
(D k y)(0) = bk , k = 0, 1, . . . , n − 1,
y(0) = y0 ,
y0 1
Y =
+ α L[f (t, y(t))],
s s
t
1
y(t) = y0 + (t − u)α−1 f (u, y(u)) du.
(α) 0
Lemma 2 (The Weierstrass Test) Suppose that {fn (t)} is a sequence of real
functions defined on a set A, and there is a sequence of positive numbers {Rn }
satisfying:
∞
∀n > 1, ∀t ∈ A, |fn (t)| ≤ Rn , Rn < ∞.
n=1
∞
Then the series fn (t) is convergent.
n=1
Theorem 1 (Existence and Uniqueness for the Caputo Problem) Let a Caputo
FDE be
y(0) = y0 .
1 V. Voltera (1860–1940).
2 P.L. Chebyshev (1821–1894).
www.pdfgrip.com
4.1 The Existence and Uniqueness Theorem for Initial Value Problems 49
on which f satisfies:
– f (t, y) is continuous,
– |f (t, y)| < M, where M = max |f (t, y)|, and Maximum (max) denotes the
(t,y)∈D
maximum function
– f (t, y) satisfy in D the Lipschitz3 condition in y if there is a constant K such
that:
1/α
η(α + 1)
δ = min η, ,
M
If we assume
|yn−1 − y0 | ≤ η,
equal with:
n−1
y0 + (yk+1 (t) − yk (t)) = yn+1 (t).
k=0
We have:
1 t
|y2 − y1 | = (t − u)α−1 [f (u, y1 (u)) − f (u, y0 )] du
(α) 0
K t KM
≤ (t − u)α−1 |y1 − y0 |du ≤ tα
(α) 0 (α + 1)
KM
≤ δ α = Kη.
(α + 1)
1 t
|y3 − y2 | = (t − u)α−1 [f (u, y2 (u)) − f (u, y1 )]du
(α) 0
K t K 2η
≤ (t − u)α−1 |y2 − y1 |du ≤ tα
(α) 0 (α + 1)
K 2η K 2 η2
≤ δα = ,
(α + 1) M
K n ηn
|yn+1 − yn | ≤ .
M n−1
www.pdfgrip.com
4.1 The Existence and Uniqueness Theorem for Initial Value Problems 51
M
The series are convergent for η < .
K
Thus the sequence {yn (t)} is uniform convergent on the compact [0, η].
Hence, yn (t) is convergent to a function y(t) for t ∈ [0, η].
∀η > 0, ∃N positive number so for n > N we have:
U = {y ∈ C[0, η] : y − y0 ≤ η}
and an operator A:
t
1
Ay(t) = y0 + (t − u)α−1 f (u, y(u)) du,
(α) 0
where A has a fixed point, and U is a closed and convex subset of all continuous
functions on [0, η] equipped with Chebyshev norm [12].
Generally:
n t
bj 1 f (u, y(u))
y(t) = (t − a)α−j + du,
(α − j + 1) (α) 0 (t − u)1−α
j =0
n
bk
y0 (t) = t α−k ,
(α − k + 1)
k=1
www.pdfgrip.com
t
1
yi (t) = y0 (t) + (t − τ )α−1 f (τ, yi−1 (τ ))dτ, i = 1, 2, . . .
(α) 0
Solution
– For α = 1, we have method of successive approximation or the method of
Picard.4
We construct a sequence {yn (t)} by the recurrence
t
yn (t) = y0 + f [u, yn−1 (u)]du, n = 1, 2, . . .
0
MN n−1 n ∂f
|y(t) − yn (t)| ≤ h , N = max .
n! (t,y)∈D ∂y
For
1
M = 2, a = 1, h = ,
2
it results:
y0 (t) = 0,
t
t3
y1 (t) = (u2 + y02 )du = ,
0 3
4 E. Picard (1856–1941).
www.pdfgrip.com
4.1 The Existence and Uniqueness Theorem for Initial Value Problems 53
t t3 t7
y2 (t) = + ,
(u2 + y12 )du =
3 63
0 t
t3 t7 2t 1 1 t 15
y3 (t) = (u + y1 )du =
2 2
+ + + ,
0 3 63 2079 59535
3
2 1 1
|y3 (t) − y(t)| ≤ 22 = , N = max |2y| = 2.
3! 2 6
y0 = 0,
t
1
yn (t) = (t − u)α−1 [u2 + yn−1
2
(u)]du.
(α) 0
2t α+2
y1 (t) = .
(α + 3)
with:
2 4 (2α + 5)
Y2 (s) = + ,
s α+3 2 (α + 3) s 3α+5
and finally:
t3 t7
+ .
For α = 1, we obtain y2 (t) =
3 63
Example 2 Using method of successive approximation we solve the FDE:
1 (2α + 1) 1 α+1
Y2 = (α + 1) + + 2α+2 ,
s α+1 (α + 1) s
2 3α+1 s
tα (2α + 1) t 3α t 2α+1
y2 = + 2 + (α + 1) .
(α + 1) (α + 1) (3α + 1) (2α + 1)
y (k) (0) = bk , k = 0, 1, 2, . . ., n − 1.
Proof The proof used here will be based on the proof of the existence and
uniqueness of the solution of Voltera integral equation.
Theorem 3 The linear FDE:
and
y (k) (0) = 0, k = 0, 1, 2, . . . , n − 1,
L[y(t)] = Y = Y (s),
L D α y(t) = L[f (t)] = F (s),
F (s)
s α Y = F (s) ⇒ Y =
sα
and using the convolution theorem it results the assumption of this theorem.
Theorem 4 The linear FDE:
y (k) (0) = bk , bk ∈ R, k = 0, 1, 2, . . . , n − 1,
n−1
y(t) = bk t k Eα,k+1 (λt α ).
k=0
L[y(t)] = Y = Y (s),
L D α y(t) = λL[y(t)],
www.pdfgrip.com
n−1
sα Y − s α−k−1 y (k) (0) − λY = 0,
k=0
n−1
n−1 α−k−1
s
n−1
Y = bk = L bk t Eα,k+1 (λt ) = L
k α k α
bk t Eα,k+1 (λt ) .
sα − λ
k=0 k=0 k=0
y(0) = A,
and where:
∞
f (n) (0)
f (t) = t n,
n!
n=0
L[y(t)] = Y = Y (s),
L D α y(t) = s α Y − As α−1 ,
∞
f (n) (0)
s α Y − As α−1 = L[f (t)] = L[t n ],
n!
n=0
∞
∞
y(t) = yn , Y = L[y(t)] = Yn ,
n=0 n=0
∞
∞
A 1 f (n) (0)
Yn = + α L[t n ],
s s n!
n=0 n=0
www.pdfgrip.com
A
Y0 = ⇒ y0 = A,
s
y(0) = A,
where:
∞
f (n) (0)
f (t) = t n,
n!
n=0
L[y(t)] = Y = Y (s),
L D α y(t) = s α Y − As α−1 ,
∞
f (n) (0)
s α Y − As α−1 + bY = L[f (t)] = L[t n ],
n!
n=0
∞
∞
y(t) = yn , Y = L[y(t)] = Yn ,
n=0 n=0
∞
∞
As α−1 1 f (n) (0)
Yn = α + α L[t n ],
as + b as + b n!
n=0 n=0
www.pdfgrip.com
As α−1 A 1
Y0 = ⇒ Y0 = ,
as α + b as b
1+
as α
∞
b f (n) (0) n b
y(t) = AEα,1 − t α + t 1 − Eα,n+1 − t α .
a a a
n=0
D α y(t) + y(t) = 1,
y(0) = 0.
www.pdfgrip.com
1
s α Y − s α−1 y(0) + Y = ,
s
1
Y = ,
s(s α+ 1)
1 1
Y = ,
s α+1 1
1+
sα
and using the identity
1
= 1 − u + u2 − u3 + . . . |u| < 1,
1+u
we obtain:
1 1 1
Y = − 2α+1 + 3α+1 + . . .
s α+1 s s
Finally, it results:
tα t 2α t 3α
y(t) = − + + ...
(α + 1) (2α + 1) (3α + 1)
MATHEMATICA
Clear["‘*"]c
ec := y’[t] + y[t] == 1;
sol = DSolve[{ec, y[0] == 0}, y, t]
Series[y[t] /. sol, {t, 0, 10}]
t2 t3 t4
y(t) = t − + − + ...
3 6 24
www.pdfgrip.com
D α y(t) = y(t),
y(0) = 0, y (0) = 1.
s α−2
Y = ,
sα − 1
1 1
Y = ,
s2 1
1−
sα
and using the identity
1
= 1 + u + u2 + u3 + |u| < 1,
1−u
we have
1 1 1
Y = + α+2 + 2α+2 + . . .,
s2 s s
and the solution:
t t α+1 t 2α+1
y(t) = + + + ...
(2) (α + 2) (2α + 2)
t t3 t5
lim y(t) = + + . . . = sinh(t).
α→2 1! 3! 5!
www.pdfgrip.com
s α−2 s α−3
Y = + b ,
sα − 1 sα − 1
s+b
lim Y = ,
α2 s(s 2 − 1)
(b − 1)e−t
r2 = Res(Y est ) = lim (s + 1)Y est = ,
−1 s→−1 2
(b + 1)et
r3 = Res(Y est ) = lim (s − 1)Y est = .
1 s→1 2
Observation
This equation can be solved also in terms of perturbation method. In this case we
take
we have
(α + 2) (2α + 2) α+1 (3α + 2) 2α+1
c1 t + c2 t + c3 t + ...
(2) (α + 2) (2α + 2)
= t + c1 t α+1 + c2 t 2α+1 + c3 t 3α+1 + . . .
(2α + 2) 1
c2 = c1 ⇒ c2 = ,
(α + 2) (2α + 2)
(3α + 2) 1
c3 = c2 ⇒ c3 = ,
(2α + 2) (3α + 2)
...
For α = 2, we have:
y (t) = y(t).
s 2 Y − sy (0) − y(0) = Y,
1
Y = ,
s2 − 1
est e−t
r1 = Res Y est = lim Y est = lim (s + 1) =− ,
−1 s→−1 s→−1 (s − 1)(s + 1) 2
est et
r2 = Res Y est = lim Y est = lim (s − 1) = ,
1 s→1 s→1 (s − 1)(s + 1) 2
et − e−t
f (t) = r1 + r2 = = sinh(t).
2
We can use here the MAPLE commands:
with(inttrans);
ec:=diff(y(t),t$2) = y(t);
dsolve(ec,D(y)(0) = 1,y(0) = 0,y(t), method = laplace);
www.pdfgrip.com
y(0) = y (0) = 1.
s2 − s − 1
= √ ,
s
L[y(t)] = Y,
1
L[t] = ,
s2
1
L[1] = ,
s
s2Y − s − 1 2
s2Y − s − 1 − √ − s Y − s − 1 = 0,
s
1 1
s 2 Y − s − 1 = 0, ⇒ Y = + 2,
s s
d 1/2 y y
= , y(0) = 0.
dx 1/2 x
which can be rewritten as:
d 1/2
L tD 1/2 y(t) = − s Y − s −1/2 y(0) − Y = 0,
ds
dY 1 1
+ + √ Y = 0.
ds 2s s
We obtain:
√
e−2 s Ce−1/t
Y (s) = C √ ⇒ y(t) = √ .
s πt
The Adomian5 method [1–3], applied to the ordinary and partial differential
equations of integer order was extended also to the case of FDE (for further details
and examples see [7, 8]).
Adomian Polynomials
We will denote these polynomials by A0 , A1 , . . . An , . . ..
We consider a nonlinear analytic function G(y(t), t) and that y(t0 ) = y0 , in
the D domain. The Adomian method consists in the decomposition the unknown
function y(t) in a series of form y(t) = y0 + y1 + y2 + . . . + yn + . . ., where yn can
be expressed in terms of Adomian polynomials An .
5 G. Adomian (1922–1996).
www.pdfgrip.com
yi → yi + (i + 1) ∗ yi+1 ∗ λ, for: i = 0, 1, . . . , k.
Step 3:
d
Ak = (k + 1) ∗ Ak .
dλ λ=0
∞
and u = un .
n=0
∞
The series An can be rearranged as a generalized Taylor series:
n=0
∞
f (u) = An = f (u0 ) + (u1 + u2 + . . .)f (1) (u0 )
n=0
u21
+ + u1 u2 + . . . f (2) (u0 ) + . . .
2!
∞
∞
(n)
= (u − u0 ) /n! f (u0 ) =
n
(u1 + u2 + . . .)n /n! f (n) (u0 ),
n=0 n=0
www.pdfgrip.com
so that:
A0 = f (u0 ),
A1 = u1 f (1) (u0 ),
...
restart;
with(LinearAlgebra):
unassign(’y,lambda’):
f:=y->y^2:
S:=lambda->sum(y[i]*lambda^i,i=0..4):
g:=lambda->(S(lambda))^2:
c:=Vector(4,n->diff(1/n!*g(lambda),lambda$n):
A:=<subs(lambda=0,g(lambda)),subs(lambda=0,c)>;
MATHEMATICA
Clear["‘*"]
f[y_] := y^2;
S[\[Lambda]_] := Sum[y[i]*\[Lambda]^i, {i, 0, 5}];
g[\[Lambda]_] := f[S[\[Lambda]]];
ad = Table[
1/n!*D[g[\[Lambda]], {\[Lambda],n}]/. \[Lambda]-> 0,
{n, 0, 5}] // Simplify;
TableForm[ad, TableAlignments -> Left]
Example 2 Let us calculate the Adomian polynomials for G = y 3 .
Solution
A0 = y03 , A1 = 3y1 y02 , A2 = 3y02 y2 + 3y12 y0 ,
A3 = y13 + 6y0 y1 y2 + 3y02 y1 , etc.
The basic principles of this algorithm remain unchanged for other definitions of the
function G.
www.pdfgrip.com
A3 = u3 (d/du0 )f (u0 ) + (u1 u2 )(d 2 /du20 )f (u0 ) + (u31 /3!)(d 3 /du30 )f (u0 ), etc.
where c is a constant.
We use the following decomposition of y(t)
∞
y(t) = yn (t),
n=0
with
∞
Ny(t) = An ,
n=0
www.pdfgrip.com
D α y(t) = t + y 2 , 1 < α ≤ 2,
y(0) = 0, y (0) = 1.
L[y(t)] = Y,
1 1
Y = + L t + y 2
,
s2 sα
so that, for the decomposition
∞
y(t) = yn (t),
n=0
www.pdfgrip.com
we obtain:
∞
∞
Y = Yn , t + y2 = An ,
n=0 n=0
∞
∞
1 1
Yn = + An ,
s2 sα
n=0 n=0
1
Y0 = ⇒ y0 = t,
s2
A0 = t + t 2 ,
1 1
Y1 = L[A0 ] ⇒ Y1 = L t + t2 ,
sα s α
1 1 t α+1 t α+2
Y1 = +2 ⇒ y1 = +2 ,
s α+2 s α+3 (α + 2) (α + 3)
A1 = 2y0 y1 ,
t α+2 t α+3
A1 = 2 +4 ,
(α + 2) (α + 3)
1 2(α + 2) 4(α + 3)
Y2 = α
L[A1 ] ⇒ Y2 = 2α+3 + 2α+4 ,
s s s
For α = 2, we obtain:
The result
t3 t4
y(t) = t + + + ...,
6 12
can be obtained also on computer, using the sequences:
MAPLE
MATHEMATICA
ec:=y"[t] == t + y[t]*y[t];
sol=DSolve[{ec,y[0]==0,y’[0]==1},y,t];
Series[y[t]/.sol,{t,0,10}]
MATHEMATICA
Clear["‘*"]
Manipulate[
f[t_] := t + t^3/6 + t^4/12 + t^5/120;
y[t_] :=
t + t^(a + 1)/Gamma[a + 2] + 2*t^(a + 2)/Gamma
[a + 3] + 2*(a + 2)*t^(2*a + 2)/Gamma[2*a + 3] +
4*(a + 3)*t^(2*a + 3)/Gamma[2*a + 4];
Plot[{f[t], y[t]}, {t, 0, 1}, ImageSize -> 300,
Frame -> True], {{a, 1/2}, 0, 1}]
The functions f (t) and y(t) calculated here are plotted in Fig. 4.1.
Example 2 Let us solve the FDE:
where:
y(0) = 0,
L[y(t)] = Y,
www.pdfgrip.com
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Fig. 4.1 Plots of the functions f (t) and y(t) from the Example 1
1 1 2
Y = + L y .
s α+1 sα
For the decomposition
∞
y(t) = yn (t),
n=0
we obtain
∞
∞
Y = Yn , y2 = An ,
n=0 n=0
∞
∞
1 1
Yn = + An ,
s α+1 sα
n=0 n=0
1 tα
Y0 = ⇒ y0 = ,
s α+1 (α + 1)
t 2α
A0 = y02 = ,
2 (α + 1)
1 1 t 2α
Y1 = L[A0 ] ⇒ Y1 = L ,
sα sα 2 (α + 1)
1 (2α + 1)
Y1 = ⇒ y1 = L−1 Y1 ,
s 3α+1 2 (α + 1)
(2α + 1) t 3α
y1 = ,
2 (α + 1) (3α + 1)
A1 = 2y0 y1 ,
1
Y2 = L[A1 ],
sα
(2α + 1) (4α + 1) 1
Y2 = 2 ,
3 (α + 1) (3α + 1) s 5α+1
y2 = L−1 [Y2 ],
(2α + 1) (4α + 1)
y2 = 2 t 5α ,
3 (α + 1) (3α + 1)(5α + 1)
tα (2α + 1) t 3α
y(t) = + 2
(α + 1) (α + 1) (3α + 1)
(2α + 1) (4α + 1)
+2 t 5α + . . .
3 (α + 1) (3α + 1)(5α + 1)
MATHEMATICA
ec:=y’[t] == 1 + y[t]*y[t];
sol=DSolve[{ec,y[0]==0},y,t];
Series[y[t]/.sol,{t,0,10}]
Finally, we have:
t3 2t 5
y(t) = t + + + ...
3 15
where:
y(0) = y α (0) = 0,
y α (0) α
L[y(t)] = Y, y0 (t) = y(0) + t = 0,
(α + 1)
we obtain
∞
∞
Y = Yn , y2 = An ,
n=0 n=0
A0 = ey0 , A1 = y1 ey0 , . . .
∞
∞
2
Yn = L An ,
s 2α
n=0 n=0
y0 = 0,
A0 = ey0 = 1,
1 2
Y1 = L[A0 ] ⇒ Y1 = ,
s 2α s 2α+1
2t 2α
y1 = ,
(2α + 1)
A1 = y1 ey0 ,
4
Y2 = ,
s 4α+1
y2 = L−1 [Y2 ],
4t 4α
y2 = ,
(4α + 1)
2t 2α 4t 4α
y(t) = 0 + + + ...
(2α + 1) (4α + 1)
0
0.0 0.2 0.4 0.6 0.8 1.0
Fig. 4.2 Plots of the functions f (t) and y(t) from the Example 3
Here, we will extend the perturbation method for the case of FDE with the aid of
some examples.
Example 1 Find the solution of the FDE:
y(0) = 0,
D α y0 (t) = 1, y0 (0) = 0,
...
1 1 tα
s α Y0 − y0 (0) = , Y0 = ⇒ y0 (t) = ,
s s α+1 (α + 1)
1 (2α + 1) (2α + 1) 1
s α Y1 = , Y1 = ,
2 (α+ 1) s 2α+1 2 (α + 1) s 3α+1
(2α + 1) t 3α
y1 (t) = .
2 (α + 1) (3α + 1)
It results also:
(2α + 1)
s α Y2 = 2 L t 4α ,
3 (α + 1)(3α + 1)
(2α + 1) (4α + 1)
Y2 = 2 ,
3 (α + 1)(3α + 1) s 5α+1
(2α + 1) (4α + 1) 5α
y2 = 2 t ,
3 (α + 1)(3α + 1) (5α + 1)
will be:
tα (2α + 1) t 3α
y(t) = + 2 +
(α + 1) (α + 1) (3α + 1)
The plot of f (t) and y(t) is done with the following program and presented in
Fig. 4.3.
MATHEMATICA
Clear["‘*"]
Manipulate[
f[t_] := t + t^3/3 + 2/15*t^5;
y[t_] :=
t^a/Gamma[a + 1] +
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Fig. 4.3 Plots of the functions f (t) and y(t) from the above Example 1
www.pdfgrip.com
−y(t)
D α y(t) = , 0 < α ≤ 1,
1+
1
= 1 − + 2 − 3 + . . .
1+
1 2 1
cos() = 1 − + 4 + . . . ,
2! 4!
which, replaced in the equation, gives:
D α y0 = −y0 , y0 (0) = 1,
D α y1 = y0 − y1 , y1 (0) = 0,
1
D α y2 = −y2 + y1 + y0 , y2 (0) = − ,
2
...
s α−1
s α Y0 − y0 (0)s α−1 = −Y0 , Y0 =
sα + 1
www.pdfgrip.com
1
L[D α y2 ] = −L[y2 ] + L[y1 ] + L[y0 ], y2 (0) = − ,
2
sα
(s α + 1)Y1 = Y0 , Y1 = ,
(s α + 1)2
s α−1 s α−1 1 sα
Y2 = + − ,
(s α + 1)3 (s α + 1)2 2 sα + 1
tα t 2α t 3α
y0 (t) = 1 − + − + ...,
(α + 1) (2α + 1) (3α + 1)
1 1
Y1 = ,
s α+1 1 2
1+ α
s
1 1 1
Y1 = −2 +3 + ...,
s α+1 s 2α+1 s 4α+1
and by inverse LT, we obtain:
tα t 2α t 4α
y1 (t) = −2 +3 + ...
(α + 1) (2α + 1) (4α + 1)
But:
s α−1 1 1 1 1
= 2α+1 2 − 2 · 3 α + 4 · 3 2α + . . . .
1 3 s 2 s s
s 3α 1+ α
s
It results also:
1 t 2α t 3α
y2 (t) = 2 −2· 3 + ...
2 (2α + 1) (3α + 1)
tα t 2α t 4α
+ −2 +3 + ...
(α + 1) (2α + 1) (4α + 1)
1 tα t 2α t 3α
− 1− + − + ... .
2 (α + 1) (2α + 1) (3α + 1)
www.pdfgrip.com
L[x(t)] = X, L[y(t)] = Y,
Example We apply the successive approximation method for the system of FDE
with initial conditions:
⎧ α
⎨ D x(t) = 3.5y(t)(1 − y(t)), x(0) = 0.2,
⎩
D α y(t) = 4x(t)(1 − x(t)), y(0) = 0.2.
We have:
⎧ t
⎪
⎪ 3.5
⎪ xn (t) = 0.2 +
⎪ yn−1 (u) − yn−1
2
(u) (t − u)α−1 du,
⎨ (α) 0
⎪
⎪ t
⎪
⎪ 4
⎩ yn (t) = 0.2 + xn−1 (u) − xn−1
2
(u) (t − u)α−1 du,
(α) 0
⎧ t
⎪
⎪ 3.5
⎪
⎪ x1 = 0.2 + 0.2 − (0.2) 2
(t − u)α−1 du,
⎨ (α) 0
⎪
⎪ t
⎪
⎪ 4
⎩ y1 = 0.2 + 0.2 − (0.2)2 (t − u)α−1 du.
(α) 0
Using the theorem regarding the product of convolution, we obtain the following
three iterations:
⎧ tα
⎪
⎪ x1 (t) = 0.2 + 0.56 ,
⎪
⎨ (α + 1)
⎪
⎪ tα
⎪
⎩ y1 (t) = 0.2 + 0.56 .
(α + 1)
⎧ t
⎪
⎪ 3.5
⎪
⎪ x2 (t) = 0.2 + y1 (u) − y21 (u) (t − u)α−1 du,
⎨ (α) 0
⎪
⎪ t
⎪
⎪ 4
⎩ y2 (t) = 0.2 + x1 (u) − x21 (u) (t − u)α−1 du,
(α) 0
www.pdfgrip.com
⎧
⎪
⎪ x2 (t) = 0.2
⎪
⎪ tα t 2α (2α + 1) t 3α
⎪
⎪
⎪
⎪ +0.56 + 1.344 − 0.808 2 ,
⎨ (α + 1) (2α + 1) (α + 1) (3α + 1)
⎪
⎪
⎪
⎪ y2 (t) = 0.2
⎪
⎪
⎪
⎪ tα t 2α (2α + 1) t 3α
⎩ +0.64 + 1.344 − 0.7077 2 ,
(α + 1) (2α + 1) (α + 1) (3α + 1)
⎧ t
⎪
⎪ 3.5
⎪
⎪ x3 (t) = 0.2 + y2 (u) − y 2
(u) (t − u)α−1 du
⎨ (α) 0 2
⎪
⎪ t
⎪
⎪ 4
⎩ y3 (t) = 0.2 + x2 (u) − x22 (u) (t − u)α−1 du
(α) 0
...
MATHEMATICA
Clear["‘*"]
Array[x, 10]
Array[y, 10]
For[{n = 0, x[0] = 0.2, y[0] = 0.2}, n < 4,
n++, {x[n + 1] =
0.2 + 3.5 InverseLaplaceTransform[
1/s^0.9 LaplaceTransform[y[n] - (y[n])^2, t, s],
s, t]// FullSimplify,
y[n + 1] =
0.2 + 4 InverseLaplaceTransform[
1/s^0.9 LaplaceTransform[(x[n] - (x[n])^2), t, s],
s, t]//
FullSimplify, Print["x=", x[n], " , ", "y= ", y[n]]}]
x = 0.2, y=0.2
x = 0.2+0.582262 t^0.9 , y= 0.2+0.665443 t^0.9
F (y) = y − y 2 .
We have also:
1
A3 = φ (0),
3! 3
...
L[x(t)] = X, L[y(t)] = Y,
∞
∞
0.2 3.5
Xn = + αL Bn
s s
n=0 n=0
∞
∞
0.2 4
Yn = + αL An
s s
n=0 n=0
0.2
X0 = ⇒ x0 = 0.2
s
www.pdfgrip.com
References 85
0.2
Y0 = ⇒ x0 = 0.2
s
⎧ 3.5 0.56 tα
⎪
⎪ X = L[B ] = , ⇒ x (t) = 0.56
⎪
⎨ 1
sα
0
s α+1
1
(α + 1)
⎪
⎪ tα
⎪
⎩ Y1 = 4 L[A0 ] = 0.64 , ⇒ y1 (t) = 0.64
sα s α+1 (α + 1)
tα
B1 = y1 − 2y0 y1 = y1 (1 − 2y0 ) = 0.64 (1 − 2 · 0.2)
(α + 1)
tα tα
= 0.64 · 0.6 · = 0.384
(α + 1) (α + 1)
tα
A1 = x1 − 2x0 x1 = x1 (1 − 2x0 ) = 0.56 (1 − 2 · 0.2)
(α + 1)
tα tα
= 0.56 · 0.6 · = 0.336
(α + 1) (α + 1)
3.5 1.344 t 2α
X2 = L[B 1 ] = ⇒ x2 (t) = 1.344
sα s 2α (2α + 1)
4 1.344 t 2α
Y2 = L[A1 ] = 2α ⇒ y2 (t) = 1.344
s α s (2α + 1)
References
4. El’sgol’ts, L. E., & Norkin, S. B. (1973). Introduction to the theory of differential equations
with deviating arguments. Mathematics in science and engineering. New York: Academic
Press.
5. Kazem, S. (2013). Exact solution of some linear differential equations by Laplace transform.
International Journal of Nonlinear Science, 16, 3–11.
6. Khan, M., Hussain, M., Jafari, H., & Khan, Y. (2010). Application of Laplace decomposition
method to solve nonlinear coupled partial differential equations. World Applied Sciences
Journal, 9, 13–19.
7. Khelifa, S., & Cherruault, Y. (2000). New results for the Adomian method. Kybernetes, 29,
332–354.
8. Milici, C., & Drăgănescu, G. (2014). A method for solve the nonlinear fractional differential
equations. Saarbrücken: Lambert Academic Publishing.
9. Natanson, I. P. (1950). Teoria funcţii vescestvennoi peremennoi. Gosudarstvennoe izdatelstvo
tehniko-teoreticekoi literaturi, Moscva.
10. Pinkus, A. (2000). Weierstrass and approximation theory. Journal of Approximation Theory,
107, 1–66.
11. O’Shaughnessy, L. (1918). Problem 433. The American Mathematical Monthly, 25, 172.
12. Weilbeer, M. (2005). Efficient numerical methods for fractional differential equations and
their analytical background. PhD thesis, Facultät für Mathematik und Informatik, Technischen
Univerisität Braunschweig, Braunschweig.
www.pdfgrip.com
Chapter 5
Generalized Systems
This chapter addresses the generalization of classical models and systems in the
perspective of FC. The following sections study the Cornu, Emden, Hermite,
Legendre, and Bessel fractional systems.
These functions can be represented in Mathematica with the aid of the following
program for q = 1/2 (see Fig. 5.1).
MATHEMATICA
Clear["‘*"]
q = 1/2;
u[t_] = 1/Gamma[q]*Integrate[(t-s)^(q-1)*Cos[s^2],
{s, 0, t}];
88 5 Generalized Systems
1.0
0.5
0.5
0.0
0.0
–0.5
–0.5
–1.0
0 2 4 6 8 10 0 2 4 6 8 10
Fig. 5.1 The graph of fractional cosine (left) and sine (right) for q = 1/2
We will introduce the following Cornu1 fractional differential system with initial
conditions:
⎧ q
⎨ D x(t) = cos(t 2 ), x(0) = 0,
⎩
D q y(t) = sin(t 2 ), y(0) = 0.
⎪
⎪
⎪
⎩ s q Y = 2! − 6! + 10! − 14! + . . . ,
s3 3!s 7 5!s 11 7!s 15
⎧
⎪ 1 4! 8! 12!
⎪
⎨ X = s q+1 − 2!s q+5 + 4!s q+9 − 6!s q+13 + . . . ,
⎪
⎪
⎪
⎪
⎩ Y = 2! − 6! + 10! − 14! + . . .
s q+3 3!s q+7 5!s q+11 7!s q+15
The graph of this curve can be written in Maple and Mathematica as:
MAPLE
restart;a:=1/2:
> x(t):=1/GAMMA(a)*int((t-u)^(a-1)*cos(u^2),u=0..t):
> y(t):=1/GAMMA(a)*int((t-u)^(a-1)*sin(u^2),u=0..t):
> plot([x(t),y(t),t=0..10],color=black,
scaling=constrained);
MATHEMATICA
Clear["‘*"]
q = 1/2;
u[t_] = 1/Gamma[q]*Integrate[(t - s)^(q - 1)*Cos[s^2],
{s, 0, t}];
v[t_] = 1/Gamma[q]*Integrate[(t-s)^(q-1)*Sin[s^2],
{s, 0, t}];
www.pdfgrip.com
90 5 Generalized Systems
Clear["‘*"]
Manipulate[
ParametricPlot[{1/Gamma[q]
*Integrate[(t - u)^(q - 1)*Cos[u^2], {u, 0, t}],
1/Gamma[q]*Integrate[(t - u)^(q - 1)*Sin[u^2],
{u, 0, t}]}, {t, 0, 10}, Frame -> True,
ImageSize -> 300], {{q, 3/2}, 0, 2}]
This curve can be plotted (see Fig. 5.2) in a plane using Maple and Mathematica as:
MAPLE
> restart;
> with(plots):
> a:=1/2:
> x:=1/GAMMA(a)*int((t-u)^(a-1)*cos(u^2),u=0..t):
> y:=1/GAMMA(a)*int((t-u)^(a-1)*sin(u^2),u=0..t):
> spacecurve([x,y,3 - x - y,t = 0..4],color=black,
scaling=constrained);
MATHEMATICA
Clear["‘*"]
Manipulate[
{x[t], y[t]} = {Integrate[Cos[u^2], {u, 0, t}],
Integrate[Sin[u^2], {u, 0, t}]};
p = ({x, y, z} /. First@Solve[x/a + y/b + z/c - 1
== 0, z]);
www.pdfgrip.com
a
b
c
1.0
0.5
1.0
0.0
–0.5
–1.0
0.0 0.5
0.5
0.0
1.0
Projection of the Cornu fractional curve on sphere (see Fig. 5.3) can be obtained in
Maple and Mathematica as:
www.pdfgrip.com
92 5 Generalized Systems
MAPLE
> restart;
> with(plots):
> a:=1/2:
> x:=1/GAMMA(a)*int((t-u)^(a-1)*cos(u^2),u=0..t):
> y:=1/GAMMA(a)*int((t-u)^(a-1)*sin(u^2),u=0..t):
> spacecurve([cos(x)*sin(y),sin(x)*sin(y),cos(y),
t=0..4],
color=black,scaling=constrained);
MATHEMATICA
Clear["‘*"]
p = ParametricPlot3D[{Cos[u]*Sin[v], Sin[u]*Sin[v],
Cos[v]},
{u, 0, 2*Pi}, {v, 0, Pi}, PlotStyle->{Cyan,
Opacity[0.3]},
Mesh -> None, Axes -> False, Boxed -> False];
q = 1/2;
u[t_] = 1/Gamma[q]*Integrate[(t - s)^(q - 1)*Cos[s^2],
{s, 0, t}];
v[t_] = 1/Gamma[q]*Integrate[(t - s)^(q - 1)*Sin[s^2],
{s, 0, t}];
q1 = ParametricPlot3D[{Cos[u[t]]*Sin[v[t]],
Sin[u[t]]*Sin[v[t]], Cos[v[t]]}, {t, 0, 10},
PlotStyle -> {Red, Opacity[0.3]},
www.pdfgrip.com
This projection (see Fig. 5.4) can be obtained with Mathematica as:
Clear["‘*"]
x[u_, v_] := u*Cos[v];
y[u_, v_] := u*Sin[v];
z[u_, v_] := -u;
Manipulate[
u[t_] = 1/Gamma[q]*Integrate[(t - s)^(q - 1)*Cos[s^2],
{s, 0, t}];
v[t_] = 1/Gamma[q]*Integrate[(t - s)^(q - 1)*Sin[s^2],
{s, 0, t}];
p = ParametricPlot3D[{x[u, v], y[u, v], z[u, v]}, {
u, -1, 1}, {v, 0, 2*Pi}, Mesh -> None, Axes
-> False, Boxed -> False];
q1 = ParametricPlot3D[{x[u[t], v[t]], y[u[t], v[t]],
z[u[t], v[t]]}, {t, 0, 10}, Mesh -> None,
Axes -> False, Boxed -> False];
Show[p, q1], {{q, 1/2}, 0, 2}]
94 5 Generalized Systems
We will use here the power series method [3] to solve the FDE. We denote by:
t t
1 α
f (t)(dt) =
α
(t − u)α−1 f (u)du = I α f (t),
(α + 1) 0 (α + 1 0
the FI of order α of f (t). Then, the norm of order α on LL2 of f (t) can be defined:
[0,1]
1 1/2
1
I α f (t)αL2 = f 2 (t)(dt)α ) .
[0,1] (α + 1) 0
For details, regarding this theorem,2 the reader can see [1, 7, 8].
We denote by C[a, b] the set of all continuous functions on [a, b].
Definition 4 For a function f ∈ C[a, b] the norm is defined
f = max |f (t)|,
t∈[a,b]
is dense in A if:
A = X.
2 H. Müntz (1884–1956).
www.pdfgrip.com
where
p(t) ∈ (),
1
Er(t q , p(t))L2 [0,1]) = I [C1 , C2 , . . . , Cn ] = [t q − p(t)]2 dt
0
2
1
n
= tq − Ck t αk dt → min,
0 k=1
2
1
n
I [C1 , C2 , . . . , Cn ] = t − q
Ck t αk
dt → min .
0 k=1
∂I [C1 , C2 , . . . , Cn ]
= 0, i = 1, 2, . . . n.
∂Ci
96 5 Generalized Systems
1
n
|q − αk |
Er(t q , p(t))L2 [0,1] = √ .
2q + 1 k=1 q + αk + 1
Q0 (t) = t q ,
...
1
Qn (t) = (αn − q)t αn Qn−1 (u)u−(1+αn ) du.
t
n−1
Qn (t) = t q − Ck t αk .
k=0
1
n−2
Qn (t) = (αn − q)t αn u−(1+αn ) uq − Ck uαk du,
t k=0
n−2
Ck
Qn (t) = t q − t αn + (αn − q) (t αk − t αn ),
αn − αk
k=0
Q0 (t) = 1,
n
q
Qn (t) ≤ 1 − ,
αk
k=1
n q
ln Qn (t) = ln 1 − .
αk
k=1
98 5 Generalized Systems
in order to write the following Maple application for calculation of the aij elements.
MAPLE
with(Student[LinearAlgebra]):
Digits:=4:
A:= 5/7, 10/17, 20/39 > | < 10/17, 1/2, 4/9 > | < 20/39, 4/9, 2/5 ;
b:= 15/23 > | < 6/11 > | < 12/25 ;
C:= evalf(LeastSquares(A,b));
f(t):=tˆ(1/3);
g(t):=0.3335*tˆ(1/5) + 0.9676*tˆ(1/2) - 0.3027*tˆ(3/4);
plot(f(t),g(t),t=0..1);
www.pdfgrip.com
G:=abs(f(t) - g(t));
for t from 0 to 1 by 0.2 do G od;
It results:
Error of Type α
a32 = a23
1 (3/2 + 1)
a33 = t 3/2 (dt)α =
0 (3/2 + 1 + α)
1 (1/3 + 1/5 + 1)
b1 = t 1/3+1/5 (dt)α =
0 (1/3 + 1/5 + 1 + α)
1 (1/3 + 1/2 + 1)
b2 = t 1/3+1/2 (dt)α =
0 (1/3 + 1/2 + 1 + α)
www.pdfgrip.com
1 (1/3 + 3/4 + 1)
b3 = t 1/3+3/4 (dt)α =
0 (1/3 + 3/4 + 1 + α)
AC T = bT ⇒ C T = A−1 bT
MAPLE
restart;
Digits: = 4;
a11:= evalf(GAMMA(2/5 + 1)/GAMMA(2/5 + 1 + a));
a12:= evalf(GAMMA(1/5 + 1/2 + 1)/GAMMA(1/5 + 1/2 + 1 + a);
a13:= evalf(GAMMA(1/5 + 3/4 + 1)/GAMMA(1/5 + 3/4 + 1 + a);
a21:= a12;
a22:= evalf(A(2 + a))
a23:= evalf(1/2 + 3/4 + 1)/GAMMA(1/2 + 3/4 + 1 + a))
a31:= a13;
a32:= a23;
a33:= evalf(3/2 + 1)/GAMMA(3/2 +1 + a))
b1:= evalf(1/3 + 1/5 + 1)/GAMMA(1/3 + 1/5 + 1 + a))
b2:= evalf(1/3 + 1/2 + 1)/GAMMA(1/3 + 1/2 + 1 + a))
b3:= evalf(1/3 + 3/4 + 1)/GAMMA(1/3 + 3/4 + 1 + a))
ec1:= C1*a11 + C2*a12 + C3*a13 = b1;
ec2:= C1*a21 + C2*a22 + C3*a23 = b2;
ec3:= C1*a31 + C2*a32 + C3*a33 = b3;
solve({ec1,ec2,ec3},{C1,C2,C3});
for α = 2/3, C1 = 0.5444, C2 = 0.3366, C3 = 0.1264
f(t):= tˆ(1/3);
g(t):= 0.54448*tˆ(1/5) + 0.3366*tˆ(1/2) + 0.1264*tˆ(3/4);
plot(f(t),g(t),t = 0..1);
G:= abs(f(t) - g(t));
for t from 0 to 1 by 0.2 do G od;
From the last program, we obtain:
Then
t kα
k k t
Ck t = Ck t0 ≤ M .
t0 t0
∞
Then Ck t kα is absolutely convergent and therefore convergent.
k=0
∞
Remark If Ck t kα , diverges for t = t0 , then it diverges for t > t0 .
k=0
Example 1 Let us determine the solution of the FDE:
We have:
∞
∞
(nα + 1)
D (2α) y(t) = Cn D (2α) (t nα ) = Cn t (n−2)α
((n − 2)α + 1)
n=0 n=2
∞
((n + 2)α + 1) nα
= Cn+2 t .
(nα + 1)
n=0
Hence:
ω2 (nα + 1)
Cn+2 = − Cn , n = 0, 1, . . . ,
((n + 2)α + 1)
B ω2 A
C0 = A, C1 = , C2 = − , ...
(α + 1) (2α + 1)
∞
∞
(−1)n ω2n 2nα (−1)n ω2n
y(t) = A t +B t (2n+1)α ,
(2nα + 1) ((2n + 1)α + 1)
n=0 n=0
www.pdfgrip.com
or:
∞
(−1)n ω2n
y(t) = AE2α (−ω2 t 2α ) + B t (2n+1)α .
((2n + 1)α + 1)
n=0
y (i) (0) = 0, i = 0, 1, . . . , m − 1,
⎛ ⎞
∞
∞
n
(nα + 1) ⎝
Cn t (n−k−1)α − Cj Cn−j ⎠
((n − k − 1)α + 1)
n=k+1 n=k j =0
(nα + 1)
× t (n−k)α = χk ,
((n − k − 1)α + 1)
where
⎧
⎨ 1, k = 0,
χk =
⎩
0, k ≥ 1.
www.pdfgrip.com
C0 = 0,
1
C1 = ,
(α + 1)
(kα + 1) n
Ck+1 = Cj Ck−j ,
((1 + k)α + 1)
j =0
We apply now the series of powers method to solve the FDE of Lane4 and Emden5
[5]
a1 a2 an
D α y(t) + D β1 y(t) + α−β D β2 y(t) + . . . + α−β D βn y(t) + y m (t) = 0,
t α−β1 t 2 t n
y(0) = 1, y (0) = 0.
n
ai
L[y(t)] = D α y + α−β
D βi y(t).
t i
i=0
It results:
∞
∞
(αk + 1) kα−α (αk + 1 + α) kα
Dα y = Ck t = Ck+1 t
(αk + 1 − α) (αk + 1)
k=1 k=0
∞
ai ai (kα + 1)
D βi y(t) = Ck t kα−βi
t α−βi t α−βi (kα + 1 − βi )
k=1
∞
(kα + 1 + α)
= Ck+1 ai t kα .
(kα + 1 + α − βi )
k=0
∞
1
n
ai
Ck+1 (αk + 1 + α) + t αk + y m = 0.
(αk + 1) (kα + 1 + α − βi )
k=0 i=1
Then:
C0 = 1,
n
ai
C1 (α + 1) 1 + + 1 = 0,
(1 + α − βi )
i=1
Ci = 0, i = 2, 3, . . .
www.pdfgrip.com
For α = 2, β = 1, a1 = 2, ai = 0, i = 2, 3, . . .
2
y (t) + y + 1 = 0,
t
resulting the solution:
t2
y(t) = 1 − .
6
(b) The case m > 0.
Then we have the following relations for different values of m:
1. m = 1, F (α, k) + Ck = 0,
k
2. m = 2, F (α, k) + Ci Ck−i = 0, because:
i=0
∞
2 ∞
k
y (t) =
2
Ck t αk
= Ci Ck−i t αk .
k=0 k=0 i=0
k
k−i
3. m = 3, F (α, k) + Ci Cj Ck−i−j = 0, because:
i=0 j =0
∞ 3 ∞
k
k−i
y (t) =
3
Ck t αk
= Ci Cj Ck−i−j t αk ,
k=0 k=0 i=0 j =0
k−i−j −...tm−1
k
k−i
F (α, k) + ... Ci Cj . . . Ck−i−...−tm = 0.
i=0 j =0 tm =0
α = 2, β = 1, a1 = 2, ai = 0, i = 2, 3 . . . ,
2 t2
(a) m = 1, y (t) + y (t) + y(t) = 0 ⇔ y(t) = 1 − + . . . ,
t 6
2 t2 t4 sin t
(b) m = 2, y (t) + y (t) + y 2 (t) = 0 ⇔ y(t) = 1 − + + . . . = ,
t 3! 5! t
www.pdfgrip.com
2 t2 t4
(c) m = 3, y (t) + y (t) + y 3 (t) = 0 ⇔ y(t) = 1 − + + ...
t 3! 60
These results are confirmed by the following Maple program:
MAPLE
>m:= ... is introduced here
>ec:= diff(y(t),t,t) + 2/t*diff(y(t),t)+(y(t))^m = 0;
>dsolve({ec, y(0) = 1, D(y)(0) = 0}, y(t), series);
Other results are as follows:
t2 t4
(1) m = 4, y(t) = 1 − + + O(t 6 ),
6 30
t2 t4
(2) m = 5, y(t) = 1 − + + O(t 6 ),
6 24
t2 t4
(3) m = 6, y(t) = 1 − + + O(t 6 ),
6 20
t2 t4
(4) m = 10, y(t) = 1 − + + O(t 6 ),
6 12
t2 t4
(5) m = 20, y(t) = 1 − + + O(t 6 ).
6 6
Finally we will examine the following particular case:
3 2 1 1 3
y ( 2 ) (t) + y ( 2 ) (t) + 3 y ( 4 ) (t) + y m (t) = 0,
t t4
3 1 3
α= , β1 = , β2 = , a1 = 2, a2 = 1,
2 2 4
⎡
3 3 5 ⎢
⎢ 1
G(k) = F ,k = k+
2 2 2 ⎣ 3
k+1
2
⎤
2 1 ⎥
+ + ⎥ ,
3 3 3 ⎦
k+2 k+
2 2 4
∞
3
(a) m = 0, G(k)t 2 k + 1 = 0
k=0
1
C0 = 1, C1 = − = −0.19742, Ci = 0, i = 2, 3, . . . ,
G(0)
3
y(t) = 1 − 0.1971 t 2 ,
1
C0 = 1, , C1 = − = −0.19742,
G(0)
C2 G(1) + C1 = 0 ⇒ C2 = 0.0146, . . . ,
3
y(t) = 1 − 0.1971t 2 + 0.0146t 3 + . . . ,
k
(c) m = 2, Ck+1 G(k) + Ci Ck−i = 0,
i=0
C0 = 1,
∞
Theorem 9 The series Ck t kα , here C0 = 0, is convergent for 0 < t < R 1/α .
k=0
Proof We can write:
C t (n+1)α
n+1
< 1,
Cn t nα
or:
C
n
0 < tα < .
Cn+1
www.pdfgrip.com
C
n
But the convergence radius is R = lim so that, finally, we obtain:
n→∞ Cn+1
If f (t) ∈ C[0, R 1/α ), and D (nα) f (t) ∈ C(0, R 1/α ), for n = 0, 1, . . ., then:
D nα f (0)
Cn = ,
(nα + 1)
(λ + 1)
D (nα) t λ = t λ−nα ,
(λ + 1 − nα)
it results:
(2α + 1) α (3α + 1) 2α
D (α) f (t) = C1 (α + 1) + C2 t + C3 t + ...
(α + 1) (2α + 1)
For:
D (α) f (0)
t = 0, we have: C1 = .
(α + 1)
D (nα) f (0)
Cn = .
(nα + 1)
∞
f (t) = (t − t0 )nα ,
n=0
∞
(nα)
Dt f (t0 )
f (t) = 0
(t − t0 )nα .
(nα + 1)
n=0
(λ + 1) λ−α
Dα t λ = t .
(λ + 1 − α)
It results:
∞
∞
∞
Cn D (2α) t nα − 2t α Cn D (α) t nα + λ Cn t nα = 0,
n=2 n=1 n=0
∞
∞
(nα + 1) (nα + 1)
Cn t (n−2)α −2t α Cn t (n−1)α
((n − 2)α + 1) ((n − 1)α + 1)
n=2 n=1
∞
+λ Cn t nα = 0,
n=0
6 C. Hermite (1822–1901).
www.pdfgrip.com
or, finally:
∞
∞ ∞
((n + 2)α + 1) nα (nα + 1)
Cn+2 t −2 Cn t nα + λ Cn t nα = 0
(nα + 1) ((n − 1)α + 1)
n=0 n=1 n=0
λC0
For n = 0 ⇒ C1 (2α + 1) + λC0 = 0 ⇒ C1 = −
(2α + 1)
((n + 2)α + 1) 2(nα + 1)
Cm+1 = Cn −λ .
(nα + 1) ((n − 1)α + 1)
For α = 1, we obtain:
λC0
C1 = − ,
2!
2n − λ
Cn+1 = Cn ,
(n + 1)(n + 2)
1 2 1
y(t) = C0 1 − λt + λ(λ − 4)t + . . . .
3
2! 4!
1
For α = , we have
2
2λC9 λ 2 C0 2
C1 = −λC0 , C2 = , C3 = √ − λ , ...,
3 3 π
or:
∞
∞
(nα + 1) (nα + 1)
Cn t (n−2)α − Cn t nα
((n − 2)α + 1) ((n − 2)α + 1)
n=2 n=2
∞
(nα + 1)
−2 Cn t nα
((n − 1)α + 1)
n=1
∞
+λ Cn t nα = 0,
n=0
For n = 0, we have:
C0 λ
C2 (2α + 1) + λC0 = 0, ⇒ C2 = − .
(2α + 1)
For n = 1, we obtain:
(3α + 1) λ − 2(α + 1)
C3 −2C1 (α +1)+λC1 = 0, ⇒ C3 = −C1 (α +1)
(α + 1) (3α + 1)
(α(n + 2) + 1) (αn + 1)
Cn+2 −Cn
(αn + 1) (α(n − 2) + 1)
(αn + 1)
−2Cn + λCn = 0.
(α(n − 1) + 1)
www.pdfgrip.com
For α = 1
λC0
C2 = − ,
6
λ−2
C3 = −C1 ,
6
λ(λ − 6)
C 4 = C0 ,
24
...
1
For α = , we have:
2
λC0
C2 = − ,
6
√
λ− π
C3 = −2C1 ,
3
1 4
C4 = −C0 1−λ+ √ .
12 π
7 F. Bessel (1784–1846).
www.pdfgrip.com
or
∞
(λ + nα + 1)
t α D (α) y(t) = t α Cn t λ+nα−α
(λ + (n − 1)α + 1)
n=0
∞
(λ + nα + 1)
= Cn t λ+nα ,
(λ + (n − 1)α + 1)
n=0
and, finally:
∞
∞
∞
(t 2α
−p ) 2
Cn t λ+nα
= Cn t λ+(n+2)α
−p 2
Cn t λ+nα
n=0 n=0 n=0
∞
∞
= Cn−2 t λ+nα − p2 Cn t λ+nα .
n=0 n=0
For α = 1, we have:
C0 λ2 − p2 = 0,
C1 (λ + 1)2 − p2 = 0.
C2k−2
C2k = − ,
(2p + 2k)(2k)
or:
C0
C2k = (−1)k+1 .
2 · 4 · 6 · · · 2k (2p + 2) · · · (2p + 2k)
and
∞
1 (−1)k t p+2k
Jp (t) = ,
2p (p + 1) 4k k!(p + 1)(p + 2) · · · (p + k)
k=0
∞
p+2k
(−1)k t
Jp (t) = ,
k!(p + k + 1) 2
k=0
where Bessel function of first kind (Jp (t)) is called the Bessel function of first kind.
1
For α = , C0 arbitrary, C1 = 0, using Mathematica we will solve the equation:
2
λ(λ)
λ+ − p2 = 0.
1
λ+
2
It results:
∞ ∞ ∞
(nα + 1)
D α
an t nα
= an D α t nα =
((n − 1)α + 1)
n=0 n=1 n=1
∞
∞
((m + 1)α + 1) mα ((n + 1)α + 1) nα
= t = t .
(mα + 1) (nα + 1)
m=0 n=0
0 0.1 0
1 0 0.3917
2 0.9475 1.5816
3 -0.2800 1.58164
4 2.3520 -0.3807
5 -2.0508 1.42210
1.0
0.5
0.0
–2 –1 0 1 2
www.pdfgrip.com
GAMMA(n alpha + 1)
f := n -> ------------------------
GAMMA((n + 1) alpha + 1)
> a[1]:=f(0)*10*(b[0]-a[0]):
> b[1]:=f(0)*(28*a[0]-a[0]*c[0]-b[0]):
> c[1]:=f(0)*(a[0]*b[0]-8/3*c[0]):
> for n from 1 to 5 do
> a[n+1]:=f(n)*10*(b[n]-a[n]);
> b[n+1]:=f(n)*(28*a[n]-sum(a[k]*b[n-k],k=0..n)-b[n]);
> c[n+1]:=f(n)*(sum(a[k]*b[n-k],k=0..n)-8/3*c[n]);
> od;
The solutions are:
y(t) = 0.1 + 2.7451t 0.95 − 1.6192t 2·0.95 + 151.18t 3·0.95 − 524.76t 4·0.95 +
References
Chapter 6
Numerical Methods
This chapter studies several numerical methods for fractional order systems. In
the following sections variational iteration, least squares, Euler’s, and Runge–Kutta
methods are analyzed.
L[y(t)] = Y (s) = Y,
m−1
(k)
L[D α y(t)] = s α Y − y0 s α−k−1 .
k=0
1
1 + λ(s)s α = 0 ⇒ λ(s) = − .
sα
It results:
1 (k) α−k−1
m−1
1
Yn+1 = α y0 s − α L[R[yn ] + N [yn ] − g(t)],
s s
k=0
or:
1 (k) α−k−1
m−1
−1 1
yn+1 (t) = L y0 s − α L[R[yn ] + N [yn ] − g(t)] .
sα s
k=0
y(0) = 1, 0 < α ≤ 1.
L[y(t)] = Y,
t
1
s Y −s
α α−1
= +L y(u) du ,
s 0
t
1
Yn+1 = Yn + λ(s) s α Yn − s α−1 − − L y(u) du ,
s 0
δYn+1 1
= 0 ⇒ 1 + λ(s)s α = 0 ⇒ λ(s) = − α ,
δYn s
t
−1 1 1 1
yn+1 = L + α+1 + α L y(u) du ,
s s s 0
tα −1 Yn
yn (t) = 1 + +L
(α + 1) s α+1
y0 = 1,
tα t α+1
y1 = 1 + + ,
(α + 1) (α + 2)
...
L[y(t)] = Y,
1 1
s α Y − s α−1 = + 2 L [yn (u)] ,
s s
1
Yn+1 = Yn + λ(s) s α Yn − s α−1 − − L [yn ] ,
s
δYn+1 1
= 0 ⇒ 1 + λ(s)s α = 0 ⇒ λ(s) = − α ,
δYn s
1 1 1
yn+1 = L−1 + α+1 + α+2 L [yn ] ,
s s s
tα Yn
yn (t) = 1 + + L−1 α+2 ,
(α + 1) s
y0 = 1,
www.pdfgrip.com
tα t α+2
y1 = 1 + + .
(α + 1) (α + 3)
...
n
yapp = Ci φ i ,
i=1
∂I [C1 , C2 , . . . , Cn ]
= 0, i = 1, 2, . . . n.
∂Ci
D α y(t) + 1 − (1 + α)t = 0,
www.pdfgrip.com
tα t α+1
s α Y = −L[1] + (α + 1)L[t], ⇔ y(t) = − + .
(α + 1) (α + 1)
For
yapp = C1 φ1 + C2 φ2 ,
D α yapp = C1 D α φ1 + C2 D α φ2 .
we obtain:
1
1 2
I [C1 , C2 ] = C1 D α φ1 + C2 D α φ2 + 1 − (α + 1)t (dt)α → min,
(α + 1) 0
⎧ ∂I [C , C ]
⎪
⎪
1 2
= 0,
⎪
⎨ ∂C1
⎪
⎪
⎩ ∂I [C1 , C2 ] = 0,
⎪
∂C2
⎧ 1
⎪
⎪
⎪ C1 D α φ1 + C2 D α φ2 + 1 − (1 + α)t D α φ1 (dt)α = 0,
⎪
⎨ 0
⎪
⎪ 1
⎪
⎪
⎩ C1 D α φ1 + C2 D α φ2 + 1 − (1 + α)t D α φ2 (dt)α = 0,
0
where i, j = 1, 2. It follows:
⎛ ⎞
B11 B12 ,
B=⎝ ⎠, C = [C1 , C2 ], F = [F1 , F2 ].
B21 B22 ,
www.pdfgrip.com
BC T = F T ,
C T = B −1 F T .
(k + 1) k−α
Dα t k = t
(k + 1 − α)
(k + 1) k+α
I αtk = t ,
(k + 1 + α)
v1 = t (1 − t), v2 = t 2 (1 − t),
yapp = C1 v1 + C2 v2 ,
D α yapp = C1 D α v1 + C2 D α v2 .
1 2
D α v1 = D α t (1 − t) = t 1−α − t 2−α ,
(2 − α) (3 − α)
2 6
D α v2 = D α (t 2 − t 3 ) = D α t 2 − D α t 3 = t 2−α − t 3−α .
(3 − α) (4 − α)
We obtain:
D α v1 = P t 1−α − Q t 2−α ,
D α v2 = P t 2−α − R t 3−α ,
(k + 1)
0I x =
α k
,
(k + 1 + α)
we obtain:
1 (3 − 2α) (4 − 2α) (5 − 2α)
B11 = D α v1D α v1(dt)α = P 2 − 2P Q + Q2 ,
0 (3 − α) (4 − α) (5 − α)
1
F1 = − [1 − (α + 1)t][P t 1−α − Qt 2−α ](dt)α
0
(2 − α) (3 − α) (3 − α) (2 − α)
= −P +Q + P (α + 1) − Q(α + 1) ,
(2) (3) (3) (2)
1
F2 = − [1 − (α + 1)t][Qt 2−α − Rt 3−α ](dt)α
0
(3 − α) (4 − α) (4 − α) (5 − α)
= −Q +R + Q(α + 1) − R(α + 1) .
(3) (4) (4) (5)
C = [C1 C2 ],
F = [F1 F2 ],
BC T = F T ⇒ C T = B −1 F T ,
resulting C1 and C2 .
The programs in Maple and Mathematica are listed in the follow-up.
www.pdfgrip.com
MAPLE
> restart:
> Digits:= 4:
> a:= 1/2;
> P:= evalf(1/GAMMA(2 - a));
> Q:= evalf(2/GAMMA(3 - a));
> R:= evalf(6/GAMMA(4 - a));
> B11:=evalf(P^2*GAMMA(3 - 2*a)/GAMMA(3 - a)
- 2*P*Q*GAMMA(4 - 2*a)/GAMMA(4-a) +
Q^2*GAMMA(5 - 2*a)/GAMMA(5 - a));
> B12:=evalf(P*Q*GAMMA(4 - 2*a)/GAMMA(4 - a)
-P*R*GAMMA(5 - 2*a)/GAMMA(5 - a) - Q^2
*GAMMA(5 - 2*a)
/GAMMA(5 - a) + Q*R*GAMMA(6 - 2*a)/GAMMA(6 - a));
> B22:=evalf(P^2*GAMMA(5 - 2*a)/GAMMA(5 - a) -
2*Q*R*GAMMA(6 - 2*a)/GAMMA(6 - a) + R^2
*GAMMA(7 - 2*a)
/GAMMA(7 - a) );
> F1:=evalf(2/GAMMA(3 + a) - 6/GAMMA(4 + a));
> F2:=evalf(6/GAMMA(4 + a) - 24/GAMMA(5 + a));
> ec1:= C1*B11 + C2*B12 = F1;
> ec2:= C1*B12 + C2*B22 = F2;
> solve({ec1,ec2},{C1,C2});
MATHEMATICA
Clear["‘*"]
a = 1/2
P = 1/Gamma[2 - a] // N
Q = 2/Gamma[3 - a] // N
C1 = 6/Gamma[4 - a] // N
v1 = P*t - B*t^2 // N
v2 = P*t^2 - C1*t^3 // N
B11 = P^2*Gamma[3 - 2*a]/Gamma[3 - a] -
2*P*Q*Gamma[4 - 2*a]/Gamma[4 - a] +
Q^2*Gamma[5 - 2*a]/Gamma[5 - a] // N
B12 = P*Q*Gamma[4 - 2*a]/Gamma[4 - a] -
P*C1*Gamma[5 - 2*a]/Gamma[5 - a] -
Q^2*Gamma[5 - 2*a]/Gamma[5 - a] +
Q*C1*Gamma[6 - 2*a]/Gamma[6 - a] // N
B22 = Q^2*Gamma[5 - 2*a]/Gamma[5 - a] -
2*Q*C1*Gamma[6 - 2*a]/Gamma[6 - a] +
C1^2*Gamma[7 - 2*a]/Gamma[7 - a] // N
F1 = -P*Gamma[2 - a]/Gamma[2] + Q*Gamma[3 -
a/Gamma[3]] +
www.pdfgrip.com
C1 = 54.3332, C2 = −39.3793.
Example 2 Establish an approximate solution, with the aid of least squares method,
for the FDE
t
D α y(t) − y(t) = , 0 ≤ t ≤ 1,
et − 1
1 1 1
B0 = 1, B1 = − , B2 = , B4 = − , . . . , B2k+1 = 0.
2 6 30
The approximate solution is in this case:
∞
yapp = Ck t αk ,
k=0
2 J. Bernoulli (1655–1705).
www.pdfgrip.com
∞
2
1 tk
I= D yapp − yapp −
α
Bk (dt)α → min,
0 k!
k=0
∞ ∞ ∞
2
1 (αk + α + 1) αk tk
I= Ck+1 t − Ck t αk − Bk (dt)α → min,
0 (αk + 1) k!
k=0 k=0 k=0
∂I
= 0 ⇒,
∂Ck
where k = 0, 1, . . .
∞
1 (αk + α + 1) 2αk Bk k+αk
Ck+1 t − Ck t 2αk
− t (dt)α = 0, k = 0, 1, . . .
(αk + 1) k!
k=0 0
C0 = B0 = 1,
1
For α = 1, we have yapp ≈ 1 − t 2 .
4
Using the Maple sequence of commands:
MAPLE
> ec:=diff(y(t),t,t) + y(t) = t/(exp(t) -1 );
> dsolve({ec,y(0) = 1,D(y)0) = 0},y(t),series);
1
it results yapp ≈ 1 − t 3 .
12
Remark Due the fact that FI is difficult to calculate it is recommended that the
function f (t) to be represented by a series of powers. So:
∞ n
t
f (t) = et = ;
n!
n=0
, - ∞
cosh(t) (±1)n n
f (t) = = t
cos(t) (2n)!
n=0
, - ∞
sinh(t) (±1)n n
f (t) = = t
sin(t) (2n + 1)!
n=0
Example 3 Establish the solution, with the aid of least squares method, for the FDE:
4t 1/2
D (3/2) y(t) − t 3/2 y(t) = √ − t 7/2 ,
π
y(0) = 0, y (0) = 0.
Using the least squares method we can build the functional I , which will be
minimized:
1
1
4t 1/2
2
I= D yapp − t
α 3/2
yapp − √ − t 3/2 (dt)3/2 → min,
(5/2) 0 π
or:
1
1
I = (D 3/2 B0 − t 3/2 B0 ) x + (D 3/2 B1 − t 3/2 B1 ) y
(5/2) 0
2
4t 1/2
+(D 3/2 B2 − t 3/2 B2 ) z − √ − t 3/2 (dt)3/2 → min.
π
1
1
4t 1/2
2
I= Ax + By + Cz − √ − t 3/2 (dt)3/2 → min,
(5/2) 0 π
Explicitly, we have:
⎧ 1 1/2
⎪
⎪
1 4t
⎪
⎪ [Ax + By + Cz] A (dt)3/2
= √ − t 3/2
A(dt)3/2 ,
⎪
⎪ π
⎪
⎪
0 0
⎪
⎪
⎪
⎨ 1 1 1/2
4t
[Ax + By + Cz] B (dt) =
3/2
√ −t 3/2
B (dt)3/2 ,
⎪
⎪ π
⎪
⎪
0 0
⎪
⎪
⎪
⎪ 1 1 1/2
⎪
⎪ 4t
⎪
⎩ [Ax + By + Cz] C (dt) 3/2
= √ − t 3/2
C(dt)3/2 .
0 0 π
www.pdfgrip.com
with:
1 1 1
A11 = A2 (dt)3/2 , A12 = B A(dt)3/2 , A13 = C A (dt)3/2 ,
0 0 0
1 4t 1/2
F1 = √ −t 3/2
A (dt)3/2
0 π
1 1 1
A21 = A B (dt)3/2 , A22 = B 2 (dt)3/2 , A23 = C B (dt)3/2 ,
0 0 0
1 4t 1/2
F2 = √ −t 3/2
B (dt)3/2
0 π
1 1 1
A31 = C A (dt)3/2 , A32 = C B (dt)3/2 , A13 = C 2 (dt)3/2 ,
0 0 0
1 4t 1/2
F3 = √ −t 3/2
C (dt)3/2 .
0 π
We observe that:
> restart;
> Digits:= 6;
> B0:= 1;
> B1:= t;
> B2:=t^2;
> A:= fracdiff(B0,t,3/2) - t^(3/2)*B0;
www.pdfgrip.com
Expand[%]
A11 = f[3]
a12 = B*A
Expand[%]
A12 = f[4]
a13 = C1*A
Expand[%]
A13 = f[5] - 4/Pi^(1/2)*f[2] // N
A21 = A12
a22 = B^2
Expand[%]
A22 = f[5]
a23 = C*B
Expand[%]
A23 = f[6] - 4/Sqrt[Pi]*f[3] // N
A31 = A13
A32 = A23
a33 = C1^2
Expand[%]
A33 = f[7] - 8/Sqrt[Pi]*f[4] + 16/Pi*f[1] // N
f1 = (4/Sqrt[Pi]*t^(1/2) - t^(7/2))*A
Expand[%]
F1 = f[5] - 4/Sqrt[Pi]*f[2] // N;
f2 = (4/Sqrt[Pi]*t^(1/2) - t^(7/2))*B
Expand[%]
F2 = f[6] - 4/Sqrt[Pi]*f[3] // N
f3 := (4/Sqrt[Pi]*t^(1/2) - t^(7/2))*C
F3 := f[7] - 8/Sqrt[Pi]*f[4] // N
Expand[%]
e3 = f[7] - 8/Sqrt[Pi]*f[4] // N
ec1 = A11*x + A12*y + A13*z
ec2 = A21*x + A22*y + A23*z
ec3 = A31*x + A32*y + A33*z
Solve[{ec1, ec2, ec3} == {F1, F2, F3}, {x, y, z}]
The Galerkin3 method is a direct method for approximate estimation of the solution
of FDE. This method will be used here to the FDE:
D α y(t) + f (t) = 0,
y(0) = 0, y(1) = 0.
N
yapp (t) = Ci φi (t),
i=1
1
[D α yapp (t) + f (t)]φi (t)(dt)α = 0, i = 1, 2, . . . , N,
0
1 1
D α yapp (t)φi (t)(dt)α = − f (t)φi (t)(dt)α , i = 1, 2, . . . , N.
0 0
Example 1 Use the Galerkin method to find the approximate solution for the FDE:
D α y(t) − t = 0,
φ1 = t, φ2 = t 2 ,
yapp = C1 t + C2 t 2 .
www.pdfgrip.com
We obtain
D α yapp = C1 D α t + C2 D α t 2 .
(k + 1) k+α
I αtk = t ,
(k + 1 + α)
⎪
⎪
⎪
1 2 2
1
⎪
⎩ C1 t 1−α
+ C2 t 2−α
t (dt) = −
2 α
t 3 (dt)α ,
0 (2 − α) (3 − α) 0
or:
⎧ 2 (3 − α) 2 (4 − α) (3)
⎪
⎪ C1 + C2 =− ,
⎪
⎨ (2 − α) (3) (3 − α) (4) (3 + α)
⎪
⎪ (4 − α) (5 − α)
⎪
⎩ C1 2 2 (4)
+ C2 =− .
(2 − α) (4) (3 − α) (5) (4 + α)
For:
α = 0.50, ⇒ C1 = −0.4156, C2 = 0.02536.
α = 1, ⇒ C1 = −0.3871, C2 = 0.08067.
Example 2 Establish an approximate solution for the following FDE, using the
Galerkin method:
D 2α y(t) + t = 0,
v1 = t 2 (1 − t), v2 = t 2 (1 − t 2 ),
www.pdfgrip.com
yapp = C1 v1 + C2 v2 .
We obtain
D 2α yapp = C1 D 2α v1 + C2 D 2α v2 ,
(k + 1)
D 2α t k = t k−2α ,
(k + 1 − 2α)
⎪
⎪
⎪
⎪
1 1 1
⎩ C1 D 2α v1 v2 (dt)α + C2 D 2α v2 v2 (dt)α = − tv2 (dt)α .
0 0 0
2 6 24
A= , B= , C= ,
(3 − 2α) (4 − 2α) (5 − 2α)
we have:
D 2α v1 = D 2α t 2 (1 − t) = At 2−2α − Bt 3−2α ,
D 2α v2 = D 2α (t 2 − t 4 ) = At 2−2α − Ct 4−2α ,
1
B11 = (D 2α v1 )v1 (dt)α
0
(5 − 2α) (6 − 2α) (6 − 2α) (7 − 2α)
=A −B −A +B ,
(5 − α) (6 − α) (6 − α) (7 − α)
1
B12 = (D 2α v1 )v2 (dt)α
0
(5 − 2α) (6 − 2α) (7 − 2α) (8 − 2α)
=A −B −A +B ,
(5 − α) (6 − α) (7 − α) (8 − α)
1
B21 = (D 2α v1 )v2 (dt)α =
0
C = [C1 C2 ],
F = [F1 F2 ],
BC T = F T ⇒ C T = B −1 F T .
www.pdfgrip.com
- A*Gamma[7 - 2*a]/Gamma[7 - a]
+ B*Gamma[8 - 2*a]/Gamma[8 - a]//N
B21 = A*Gamma[5 - 2*a]/Gamma[5 - a]
- A*Gamma[6 - 2*a]/Gamma[6 - a]
- C*Gamma[7 - 2*a]/Gamma[7 - a]
+ C*Gamma[8 - 2*a]/Gamma[8 - a]//N
B22 = A*Gamma[5 - 2*a]/Gamma[5 - a]
- A*Gamma[7 - 2*a]/Gamma[7 - a]
- C*Gamma[7 - 2*a]/Gamma[7 - a]
+ C*Gamma[9 - 2*a]/Gamma[9 - a]//N
F1 = -Gamma[4]/Gamma[4 + a] + Gamma[5]/
Gamma[5 + a]//N
F2 = -Gamma[4]/Gamma[4 + a] + Gamma[6]/
Gamma[6 + a]//N
ec1 = C1*B11 + C2*B12
ec2 = C1*B12 + C2*B22
Solve[{ec1,ec2}=={F1,F2},{F1,F2}]//N
For:
α = 0.50, C1 = −12.89, C2 = 8.135
α = 1, C1 = 1.067, C2 = −0.3944
Example 3 Establish an approximate solution using the Galerkin method, for the
FDE:
tα (2α + 1)
Ya = − + 2 t 3α + . . .
(α + 1) (α + 1)(3α + 1)
Ya = At 1/2 + Bt 3/2 ,
and we have:
⎧ 1 1
⎪
⎪
⎪
⎪ (D 1/2
Y a − Y 2 1/2
)t (dt) 1/2
= − t 1/2 (dt)1/2 ,
⎨ 0 a
0
⎪
⎪ 1
⎪
⎪
1
⎩ (D Ya − Ya )t (dt) = −
1/2 2 3/2 1/2
t 3/2 (dt)1/2 .
0 0
www.pdfgrip.com
Ex22 = Expand[%];
ec1 = 1/2*A*Sqrt[Pi]*f[1/2] + 3/4*B*Sqrt[Pi]*f[3/2]
- A^2*f[3/2] - 2 A*B*f[5/2] - B^2*f[7/2] // N;
ec2 = 1/2*A*Sqrt[Pi]*f[3/2] + 3/4*B*Sqrt[Pi]*f[5/2]
- A^2*f[5/2] - 2 A*B*f[7/2] - B^2*f[9/2] // N;
NSolve[{ec1, ec2} == {f1, f2}, {A, B}]
We introduce here a generalization of the Euler method to the case of FDE of type:
We assume that y(t), D α y(t), D 2α y(t) are continuous on [t0 , a] and we will
apply the generalized Taylor’s formula:
tα t 2α
y(t) = y(t0 ) + D α y(t0 ) + D 2α (η) , t0 < η ≤ t.
(α + 1) (2α + 1)
hα
y(t) ≈ y(t0 ) + f (t0 , y(t0 )),
(α + 1)
hα
yn+1 ≈ yn + f (tn , yn ),
(α + 1)
2t
D 1/2 y(t) = y(t) − , with the initial condition: y(0) = 1,
y(t)
> d:=0.2:
> a:=1/2:
> h:=evalf(d^a/GAMMA(a+1)):
> t:=array[0..10]:
> y:=array[0..10]:
> for k from 0 to 5 do t[k]:=k*0.2 od:
> y[0]:=1:
> for k from 0 to 5 do y[k+1]:=y[k]+h*f(t[k],y[k]) od:
> for k from 0 to 5 do print(t[k]," " ,y[k]) od;
0.0, 1
0.2, 1.504
0.4, 2.128
0.6, 3.012
0.8, 4.331
1.0, 6.329
MATHEMATICA
Clear["*"]
\[Delta] = 0.1; \[Alpha] = 1/2;
f[t_, y_] := y - 2*t/y
values =
RecurrenceTable[{t[k + 1] == t[k] + \[Delta],
y[k + 1] ==
y[k] + \[Delta]^\[Alpha]/Gamma[\[Alpha] + 1]
*f[t[k], y[k]],
t[0] == 0, y[0] == 1}, {t, y}, {k, 0, 10}];
Grid[values]
ListLinePlot[values, PlotMarkers -> Automatic]
The solution y(t) is plotted in Fig. 6.1.
Example 2 Approximate the following FDE with the aid of Euler method:
y(t) − t
D α y(t) = , for the initial condition: y(0) = 1,
y(t) + t
15
10
> t:=array[0..10]:
> y:=array[0..10]:
> for k from 0 to 10 do t[k]:=k*0.1 od:
> y[0]:=1:
> for k from 0 to 10 do y[k+1]:=y[k]+h*f(t[k],
y[k]) od:
> for k from 0 to 10 do print(t[k]," " ,y[k]) od;
0., 1
0.1, 1.357
0.2, 1.665
0.3, 1.945
0.4, 2.206
0.5, 2.453
0.6, 2.689
0.7, 2.916
0.8, 3.135
0.9, 3.347
1.0, 3.552
MATHEMATICA
Clear["*"]
\[Delta] = 0.1; \[Alpha] = 1/2;
f[t_, y_] := (y - t)/(y + t)
values =
RecurrenceTable[{t[k + 1] == t[k] + \[Delta],
www.pdfgrip.com
y[k + 1] ==
y[k] + \[Delta]^\[Alpha]/Gamma[\[Alpha] + 1]
*f[t[k], y[k]],
t[0] == 0, y[0] == 1}, {t, y}, {k, 0, 10}];
Grid[values]
ListLinePlot[values, PlotMarkers -> Automatic]
0.0 1
0.1 1.35582
0.2 1.86466
0.3 1.94494
0.4 2.2064
0.5 2.4532
0.6 2.68972
0.7 2.91639
0.8 3.13587
0.9 3.34681
1 3.5524
In this case the solution y(t) is plotted in Fig. 6.2.
Using the eulerstep[. . . ] command, the solution can be obtained using the
Mathematica program:
MATHEMATICA
Clear["‘*"]
a = 1/2;
d = 0.1;
eulerstep[f_, {t_, y_}, h_] := {t + h,
3.0
2.5
2.0
1.5
with the initial conditions x(1) = 1, y(1) = 1, the time step h = 0.2 and t ∈ [1, 2].
k1 = f (tn , yn ),
h k1
k2 = f tn + , yn + ,
2 2
yn+1 = yn + h k2 .
Clear["‘*"]
d = 0.01
www.pdfgrip.com
a = 0.98
h = d^a/Gamma[a + 1]
Eulerlor[{x_, y_, z_}] := {x - 10*h*(x - y),
y + h*(28*x - y - x*z),
z + h*(x*y - 8/3*z)}
sol = NestList[Eulerlor, {0, 1, 0}, 1000];
p = Interpolation /@ Transpose@sol;
ParametricPlot3D[Through@p@t, {t, 1, 1000},
PlotPoints -> 100,
ColorFunction -> (Hue[4 #] &), ImageSize -> 300]
Figure 6.3 shows the 3D Lorenz attractor solution (x(t), y(t), z(t)) for α = 0.98.
In this section we will suppose that all conditions of existence and uniqueness of the
solutions are fulfilled.
www.pdfgrip.com
The second order Runge4 –Kutta5 method Second order Runge–Kutta (RK2)
introduced here is an extension of the first Euler method for approximation of the
solution of the FDE:
where:
In this case the solution can be approximated in the following discrete form:
hα K1 + K2
yn+1 = yn + ,
(α + 1) 2
where
K1 = f (tn , yn ),
hα hα
K2 = f tn + , yn + K1 .
(α + 1 (α + 1)
K1 = f (tn , yn ),
hα hα
K2 = f tn + A , yn + B K1 ,
(α + 1 (α + 1
E(h) = yn+1 − yn ,
with:
hα h2α
E(h) = yn+1 − yn = D α yn + D 2α yn + O(h3α )
(α + 1) (2α + 1)
hα h2α
= f (tn , yn ) + D 2α yn + O(h3α ),
(α + 1) (2α + 1)
hα
D α E(h) = f (tn , yn ) + D α yn + O(h2α ),
(α + 1)
and
hα hα
yn+1 − yn = C1 K1 + C2 K2 .
(α + 1) (α + 1)
hα hα
E[h] = C1 K1 + C2 K2 ,
(α + 1) (α + 1)
hα
D α E[h] = c1 K1 + c2 K2 + D α K2 ,
(α + 1)
hα hα
D α K2 = D α f tn + A , yn + B K1
(α + 1 (α + 1
For h → 0 we obtain:
D 2α E(h) = D α (D α E(h)) = C2 D α K2 + D α K2 .
It results:
2C2 Aftn (tn , yn ) + 2C2 Bf (tn , yn )fyn (tn , yn ) = ftn (tn , yn ) + f (tn , yn )fyn (tn , yn ).
1 hα
For C1 = C2 = we have the classical RK2 method, with the step .
2 (α + 1)
Example Solve with RK2 the FDE
y2
D 1/2 y(t) = t 2 + , y(0) = 1.
4
The solution can be written in Maple or Mathematica:
MAPLE
> restart;
> Digits:=3:
> a:=1/2:
> d:=0.2:
> h:=evalf(d^a/GAMMA(a+1)):
> y:=Array[0..10]:
> y[0]:=1:
> unassign(’t,y’);
> f:=(t,y)->t^2+y^2/4:
> for i from 0 to 4 do
> k1:=f(i*0.2,y[i]):
> k2:=f(i*0.2 + h,y[i]+h*k1):
> y[i+1]:=y[i]+(k1+k2)/2 od:
> for i from 0 to 5 do print(i*0.2,y[i]) od;
0 1
0.2 1.207
0.4 1.560
0.6 2.200
0.8 3.468
1 6.597
www.pdfgrip.com
MATHEMATICA
Clear["‘*"]
a = 1/2
d = 0.2
h = d^a/Gamma[a + 1]
y[0] = 1
f[t_, y_] := t^2 + y^2/4
For[i = 0, i < 5,
i++, {k1 = f[t, y] /. {t -> i*0.2, y -> y[i]},
k2 = f[t, y] /. {t -> i*0.2 + h, y -> y[i] + h*k1},
y[i + 1] = y[i] + (k1 + k2)*0.5*h}]
Table[{j*0.2, y[j]}, {j, 0, 5}] // TableForm
0 1
0.2 1.20733
0.4 1.56029
0.6 2.20011
0.8 3.469
1 6.59765
where:
hα
yn+1 = yn + (K1 + 2K2 + 2K3 + K4 ),
6(α + 1)
hα h2α
E(h) = yn+1 − yn = D α yn + D 2α yn + O(h3α )
(α + 1) (2α + 1)
hα h2α
= f (tn , yn ) + D 2α yn + O(h3α ),
(α + 1) (2α + 1)
hα
D α E(h) = f (tn , yn ) + D α yn + O(h2α ),
(α + 1)
and
hα hα hα hα
yn+1 −yn = C1 K1 +C2 K2 +C3 K3 +C4 K4 .
(α + 1) (α + 1) (α + 1) (α + 1)
Hence, we have
hα hα hα hα
E[h] = C1 K1 + C2 K2 + C3 K3 + C4 K4 ,
(α + 1) (α + 1) (α + 1) (α + 1)
where:
K1 = f (tn , yn ),
hα hα
K2 = f tn + a2 , yn + b2 K1 ,
2(α + 1) 2(α + 1)
hα hα
K3 = f tn + a3 , yn + b3 K2 ,
2(α + 1) 2(α + 1)
hα hα
K4 = f tn + a4 , yn + b4 K3 ,
(α + 1) (α + 1)
and finally:
D α E[h] = C1 K1 + C2 K2 + C3 K3 + C4 K4
hα hα
+C1 D α K1 + C 2 D α K2
(α + 1) (α + 1)
hα hα
+C3 D α K3 + C 4 D α K4 .
(α + 1) (α + 1)
www.pdfgrip.com
We take now h → 0:
1 = C1 + C2 + C3 + C4 .
⎩
C2 b2 + C3 b3 + C4 b4 = 12 ,
Remark The solution of this system is not unique. For this reason we choose as in
the RK2 case:
C1 = C4 = 1/6, C2 = C4 = 1/3
1
a1 = b1 = a4 = b4 = 1, a2 = b2 = a3 = b3 =
2
Example 1 Solve, using the RK4 algorithm, the following FDE:
y2
D α y(t) = t 2 + , y(0) = 1,
4
for the cases:
(a) α = a = 12 ,
(b) α = a = 0.98.
www.pdfgrip.com
Clear["‘*"]
RK4step[{t_, y_}] := Module[{k1, k2, k3, k4},
k1 = f[t, y];
k2 = f[t + h/2, y + h/2 k1];
k3 = f[t + h/2, y + h/2 k2];
k4 = f[t + h, y + h k3];
{t + h, y + 1/6 (k1 + 2 k2 + 2 k3 + k4)}]
f[t_, y_] := t^2 + y^2/4;
t0 = 0;
y0 = 1;
a = 1/2;
h = (0.01)^a/Gamma[a + 1];
n = 5;
rkpoints = NestList[RK4step, {t0, y0}, n];
PaddedForm[TableForm[rkpoints], {6, 4}]
0, 1,
0.1128, 1.2616,
0.2256, 1.7047,
0.3385, 2.5522,
0.4513, 4.4782,
0.5641, 10.500
www.pdfgrip.com
Solution
MATHEMATICA
Clear["‘*"];
a = 1/2; x[0] := 1;
y[0] = 0; h = (0.2)^a/Gamma[a + 1]; n = 10000;
Do[t[i] = 0.0 + (0.2)*i, {i, 0, n}]
f[t_, x_, y_] = y
g[t_, x_, y_] = -x + 0.25*(1 - x^2)*y
Do[{K1 = f[t[i], x[i], y[i]],
L1 = g[t[i], x[i], y[i]],
K2 = f[t[i] + h, x[i] + h*K1, y[i] + h*L1],
L2 = g[t[i] + h, x[i] + h*K1, y[i] + h*L1],
K3 = f[t[i] + h/2, x[i] + h*K2/2, y[i] + h*L2/2],
L3 = g[t[i] + h/2, x[i] + h*K2/2, y[i] + h*L2/2],
K4 = f[t[i] + h, x[i] + h*K3, y[i] + h*L3],
L4 = g[t[i] + h, x[i] + h*K3, y[i] + h*L3],
x[i + 1] = x[i] + (K1 + 2 K2 + 2 K3 + K4)*h/6,
y[i + 1] = y[i] + (L1 + 2 L2 + 2 L3 + L4)*h/6},
{i, 0, n}];
Do[Print[t[i], " ", x[i], " ", y[i]], {i, 0, 5}];
Figure 6.4 shows the plot of the Van der Pol solution from Example 3.
0.5
0.0
–0.5
–1.0
Solution
Clear["‘*"];
a = 0.998; x[0] = 1;
y[0] = 0; h = (0.2)^a/Gamma[a + 1]; n = 10000;
Do[t[i] = 0.0 + (0.2)*i, {i, 0, n}]
f[t_, x_, y_] := y
g[t_, x_, y_] := -x - x^3
Do[{K1 = f[t[i], x[i], y[i]], L1 = g[t[i], x[i],
y[i]],
K2 = f[t[i] + h, x[i] + h*K1, y[i] + h*L1],
L2 = g[t[i] + h, x[i] + h*K1, y[i] + h*L1],
K3 = f[t[i] + h/2, x[i] + h*K2/2, y[i] + h*L2/2],
L3 = g[t[i] + h/2, x[i] + h*K2/2, y[i] + h*L2/2],
K4 = f[t[i] + h, x[i] + h*K3, y[i] + h*L3],
L4 = g[t[i] + h, x[i] + h*K3, y[i] + h*L3],
x[i + 1] = x[i] + (K1 + 2 K2 + 2 K3 + K4)*h/6,
y[i + 1] = y[i] + (L1 + 2 L2 + 2 L3 + L4)*h/6},
{i, 0, n}]; ListPlot[Table[{x[n], y[n]},
{n, 0, 10000}], Frame -> True]
Figure 6.5 shows the plot of the fractional Duffing solution from the Example 4.
Example 5 Solve fractional system
⎧
⎪
⎪ D 1/2 x(t) = 2y, x(0) = 1,
⎪
⎪
⎪
⎨
D 1/2 y(t) = 2z, y(0) = 1,
⎪
⎪
⎪
⎪
⎪
⎩ 1/2
D z(t) = x − y, z(0) = 1.
Solution
MAPLE
> restart;
> Digits:=4:X:=array[0..5]:Y:=array[0..5]:
Z:=array[0..5]:
1.x10–10
5.x10–11
–5.x10–11
–1.x10–10
Fig. 6.5 Plot of the fractional Duffing solution from the Example 4
> unassign(’t,x,y,z’):
> f:=(t,x,y,z)->2*y:
> g:=(t,x,y,z)->2*z:
> p:=(t,x,y,z)->x - y:
> a:=1/2: x:=1:X[0]:=1: y:=1:Y[0]:=1:
> z:=1:Z[0]:=1: d:=0.2:
> h:=evalf(d^a/GAMMA(a+1)):
> for n from 0 to 5 by 1 do
> k1:=h*f(d*n,x,y,z);
> l1:=h*g(d*n,x,y,z):
> m1:=h*p(d*n,x,y,z):
> k2:=h*f(d*n+h,x+k1,y+l1,z+m1):
> l2:=h*g(d*n+h,x+k1,y+l1,z+m1):
> m2:=h*p(d*n+h,x+k1,y+l1,z+m1);
> k3:=h*f(d*n+h/2,x+k2/2,y+l2/2,z+m2/2):
> l3:=h*g(d*n+h/2,x+k2/2,y+l2/2,z+m2/2):
> m3:=h*p(d*n+h/2,x+k2/2,y+l2/2,z+m2/2):
> k4:=h*f(d*n,x+k3,y+l3,z+m3):
> l4:=h*g(d*n,x+h*k3,y+h*l3,z+h*m3):
www.pdfgrip.com
> m4:=h*p(d*n,x+h*k3,y+h*l3,z+h*m3):
> x:=x+(k1+2*k2+2*k3+k4)/6:X[n+1]:=x:
> y:=y+(l1+2*l2+2*l3+l4)/6:Y[n+1]:=y:
> z:=z+(m1+2*m2+2*m3+m4)/6:Z[n+1]:=z od:
> for n from 0 to 5 do print(d*n,X[n],Y[n],Z[n]) od;
0., 1, 1, 1
0.2, 2.688, 2.030, 1.108
0.4, 5.612, 3.458, 1.795
0.6, 10.67, 6.102, 3.458
0.8, 19.87, 11.26, 6.694
1.0, 37.08, 21.11, 12.70
MATHEMATICA
Clear["‘*"]
f[t_, x_, y_, z_] := 2*y;
g[t_, x_, y_, z_] := 2*z;
p[t_, x_, y_, z_] := x - y;
a = 1/2; d = 0.2; x[0] = 1; y[0] = 1; z[0] = 1;
tmax = 5;
h = (d)^a/Gamma[a + 1];
Do[{k1 = h f[t[n], x[n], y[n], z[n]];
l1 = h g[n*d, x[n], y[n], z[n]];
m1 = h p[n*d, x[n], y[n], z[n]];
k2 = h f[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
l2 = h g[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
m2 = h p[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
k3 = h f[n*d + h/2, x[n] + k2/2, y[n] + l2/2,
z[n] + m2/2];
l3 = h g[n*d + h/2, x[n] + k2/2, y[n] + l2/2,
z[n] + m2/2];
m3 = h p[n*d + h/2, x[n] + k2/2, y[n] + l2/2,
z[n] + m2/2];
k4 = h f[n*d + h, x[n] + k3, y[n] + l3, z[n] + m3];
l4 = h g[n*d + h, x[n] + k3, y[n] + l3, z[n] + m3];
m4 = h p[n*d + h, x[n] + k3, y[n] + l3, z[n] + m3];
x[n + 1] = x[n] + 1/6 *(k1 + 2 k2 + 2 k3 + k4);
y[n + 1] = y[n] + 1/6* (l1 + 2 l2 + 2 l3 + l4);
z[n + 1] = z[n] + 1/6* (m1 + 2 m2 + 2 m3 + m4)};
Print[0.2*n, " ", x[n], " ", y[n], " ", z[n]],
{n, 0, tmax}]
www.pdfgrip.com
Example 6 (Lorenz Attractor with Interpolation Solution) Solve the Lorenz attrac-
tor problem:
⎧
⎪
⎪D
0.998 x(t) = 10(y − x), x(0) = 1,
⎪
⎪
⎪
⎨
D 0.998 y(t) = 28x − y − xz, y(0) = 1,
⎪
⎪
⎪
⎪
⎪
⎩ D 0.998 z(t) = xy − 8 z,
3 z(0) = 1,
Clear["‘*"]
f[t_, x_, y_, z_] := 10*(y - x);
g[t_, x_, y_, z_] := 28*x - y - x*z;
p[t_, x_, y_, z_] := x*y - 8/3* z;
a = 0.998; d = 0.01;
x[0] = 1;
y[0] = 1;
z[0] = 1;
tmax = 10000;
h = (d)^a/Gamma[a + 1]; Do[{k1 = h f[t[n], x[n],
y[n], z[n]];
l1 = h g[n*d, x[n], y[n], z[n]];
m1 = h p[n*d, x[n], y[n], z[n]];
k2 = h f[n*d + h/2, x[n] + k1/2, y[n] + l1/2, z[n]
+ m1/2];
l2 = h g[n*d + h/2, x[n] + k1/2, y[n] + l1/2, z[n]
+ m1/2];
m2 = h p[n*d + h/2, x[n] + k1/2, y[n] + l1/2, z[n]
+ m1/2];
k3 = h f[n*d + h/2, x[n] + k2/2, y[n] + l2/2, z[n]
+ m2/2];
l3 = h g[n*d + h/2, x[n] + k2/2, y[n] + l2/2, z[n]
+ m2/2];
m3 = h p[n*d + h/2, x[n] + k2/2, y[n] + l2/2, z[n]
+ m2/2];
k4 = h f[n*d + h, x[n] + k3, y[n] + l3, z[n] + m3];
l4 = h g[n*d + h, x[n] + k3, y[n] + l3, z[n] + m3];
m4 = h p[n*d + h, x[n] + k3, y[n] + l3, z[n] + m3];
x[n + 1] = x[n] + 1/6 (k1 + 2 k2 + 2 k3 + k4);
y[n + 1] = y[n] + 1/6 (l1 + 2 l2 + 2 l3 + l4);
z[n + 1] = z[n] + 1/6 (m1 + 2 m2 + 2 m3 + m4);},
{n,0,tmax}];
sos = Table[{x[t], y[t], z[t]}, {t, 0, tmax}];
ListPointPlot3D[sos, ImageSize -> 300]
www.pdfgrip.com
p = Interpolation /@ Transpose@sos;
ParametricPlot3D[Through@p@t, {t, 0, 10000},
PlotPoints -> 100,
ColorFunction -> (Hue[#4] &), ImageSize -> 300]
Figures 6.6 and 6.7 show the Lorenz system without and with interpolation,
respectively.
Example 7 (Rössler Attractor) Solve the fractional Rössler8 attractor system, using
the RK4 method:
⎧
⎪
⎪ D 0.98 x(t) = −y − z, x(0) = 1,
⎪
⎪
⎪
⎨
D 0.98 y(t) = x + 0.2y, y(0) = 1,
⎪
⎪
⎪
⎪
⎪
⎩ 0.98
D z(t) = 0.2 + z(x − 8), z(0) = 1.
40
30
20 20
10
0
0
–10
10 –20
20
8 O.E. Rössler(1940–).
www.pdfgrip.com
20
20
40
30
20
10
10
0
10
20
Clear["‘*"]
f[t_, x_, y_, z_] := -y - z;
g[t_, x_, y_, z_] := x + 0.2*y;
p[t_, x_, y_, z_] := 0.2 + z*(x - 8);
a = 0.998; d = 0.01; x[0] = 1;
y[0] = 1; z[0] = 1; tmax = 2000;
h = (d)^a/Gamma[a + 1];
Do[{k1 = h f[t[n], x[n], y[n], z[n]];
l1 = h g[n*d, x[n], y[n], z[n]];
m1 = h p[n*d, x[n], y[n], z[n]];
k2 = h f[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
l2 = h g[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
m2 = h p[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
www.pdfgrip.com
Figure 6.8 shows the 3D Rössler attractor solution (x(t), y(t), z(t)).
0.10
10
0.05
5
0.00
5 0
5
5
Clear["‘*"]
f[t_, x_, y_, z_] := -x - 5*y - z*y;
g[t_, x_, y_, z_] := - 85*x - y - x*z;
p[t_, x_, y_, z_] := 0.5*z + x*y + 1;
a = 0.998; d = 0.001; tmax = 10000;
x[0] = 8; y[0] = 2; z[0] = 1;
h = (d)^a/Gamma[a + 1];
Do[{k1 = h f[t[n], x[n], y[n], z[n]];
l1 = h g[n*d, x[n], y[n], z[n]];
m1 = h p[n*d, x[n], y[n], z[n]];
k2 = h f[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
l2 = h g[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
m2 = h p[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
k3 = h f[n*d + h/2, x[n] + k2/2, y[n] + l2/2,
z[n] + m2/2];
l3 = h g[n*d + h/2, x[n] + k2/2, y[n] + l2/2,
z[n] + m2/2];
m3 = h p[n*d + h/2, x[n] + k2/2, y[n] + l2/2,
z[n] + m2/2];
k4 = h f[n*d + h, x[n] + k3, y[n] + l3, z[n] + m3];
l4 = h g[n*d + h, x[n] + k3, y[n] + l3, z[n] + m3];
m4 = h p[n*d + h, x[n] + k3, y[n] + l3, z[n] + m3];
x[n + 1] = x[n] + 1/6 (k1 + 2 k2 + 2 k3 + k4);
y[n + 1] = y[n] + 1/6 (l1 + 2 l2 + 2 l3 + l4);
z[n + 1] = z[n] + 1/6 (m1 + 2 m2 + 2 m3 + m4);},
{n,0,tmax}]
sos = Table[{x[t], y[t], z[t]}, {t, 0, tmax}];
ListPointPlot3D[Table[{x[t], y[t], z[t]},{t,0,tmax}],
ImageSize -> 300]
sos = Table[{x[n], y[n], z[n]}, {n, 1, 10000}];
p = Interpolation /@ Transpose@sos;
ParametricPlot3D[Evaluate@Through@p@t,{t,1,10000},
ImageSize->300]
www.pdfgrip.com
Figures 6.9 and 6.10 show the fractional Volta attractor solution without and with
interpolation, respectively.
Example 9 (Chua Attractor System) Find the numeric solution of the Chua9 attrac-
tor system:
⎧
⎪
⎪ D 0.998 x(t) = 40(y − x), x(0) = 0,
⎪
⎪
⎪
⎨
D 0.998 y(t) = (28 − 40)x + 28y − xz, y(0) = 1,
⎪
⎪
⎪
⎪
⎪
⎩ D 0.998 z(t) = xy − 2z, z(0) = 0.
Clear["‘*"]
f[t_, x_, y_, z_] := 40*(y - x);
g[t_, x_, y_, z_] := x*(28 - 40) + 28*y - x*z;
50 40
100 20
0
100
50
20
0
50 40
100
40 100
20
50
0
20 0
40 50
50
100
Figures 6.11 and 6.12 show the fractional Chua attractor solution without and
with interpolation, respectively.
40
30
20 20
10
0 10
0
10
0
10
10
20 20
20
10
0
–10
–20
40
30
20
10
–10
0
10
20
tn = t0 + nh, n ∈ N.
The approximate solution of second order of this system can be written as:
hα
xn+1 = xn + (K1 + K2 ),
(α + 1)
www.pdfgrip.com
hβ
yn+1 = yn + (L1 + L2 ),
(β + 1)
K1 = f (tn , xn , yn ), L1 = g(tn , xn , yn ),
hα hα K1 hβ L1
K2 = f tn + , xn + , yn + ,
(α + 1) (α + 1) 2 (β + 1) 2
hβ hα K1 hβ L1
L2 = g t n + , xn + , yn + .
(β + 1) (α + 1) 2 (β + 1) 2
tα tα
xn+1 = xn + c1 K1 + c2 K2
(α + 1) (α + 1)
tα tα
yn+1 = yn + d1 L1 + d2 L2
(α + 1) (α + 1)
where
K1 = f (tn , xn , yn ),
L1 = g(tn , xn , yn ),
hα hα hβ
K2 = f tn + af α , xn + K1 bf α , yn + L1 bfβ ,
(α + 1) (α + 1) (β + 1)
hβ hα hβ
L2 = g t n + agβ , xn + K1 bgα , yn + L1 bgβ ,
(β + 1) (α + 1) (β + 1)
where the constants c1 , c2 , d1 , d2 , af α , bf α , agα , bgα , afβ , bfβ , agβ , and bgβ will
be established.
We define the errors E(h) and F (h) of expansions
hα h2α
E(h) = x(tn + h) − x(tn ) = D α x(tn ) + D 2α x(tn ) + O(h3α ),
(α + 1) (2α + 1)
hβ h2β
F (h) = y(tn + h) − y(tn ) = D β y(tn ) + D 2β y(tn ) + O(h3β ),
(β + 1) (2β + 1)
www.pdfgrip.com
D 2α E(h) = D α (D α E(h))
hα
= D (c1 K1 + c2 K2 ) + D c2
α α
D K2 + O(hα ),
α
(α + 1)
But:
Dα y = 0 because: α > β,
we obtain
⎧
⎪
⎪ 2c2 af α = 1,
⎪
⎪
⎪
⎨
2c2 bf α1 = 1,
⎪
⎪
⎪
⎪
⎪
⎩ 2c b
2 f α2 = 1.
The calculations will continue with the derivative. Similar computations can be
carried out in the case of RK4 method.
www.pdfgrip.com
It results:
1 hα
xn+1 = xn + (K1 + 2K2 + 2K3 + K4 ),
6 (α + 1)
1 hβ
yn+1 = yn + (L1 + 2L2 + 2L3 + L4 ).
6 (β + 1)
Example 1 (Modified Duffing System) Find the RK4 solution for the modified
fractional Duffing system:
⎧ 0.998
⎨D x(t) = y, x(0) = 1,
⎩
D 0.50 y(t) = −x + 0.25(1 − x 3 )y, y(0) = 0.
Clear["‘*"];
a = 0.998; b = 0.5; x[0] = 1; y[0] = 0;
h = (0.2)^a/Gamma[a + 1]; l = (0.2)^b/Gamma[b + 1];
n = 10000;
Do[t[i] = 0.0 + (0.2)*i, {i, 0, n}]
f[t_, x_, y_] := y
g[t_, x_, y_] := -x + 0.25*(1 - x^2)*y
Do[{K1 = f[t[i], x[i], y[i]], L1 = g[t[i], x[i],
y[i]],
K2 = f[t[i] + h, x[i] + h*K1, y[i] + h*L1],
L2 = g[t[i] + l, x[i] + l*K1, y[i] + l*L1],
K3 = f[t[i] + h/2, x[i] + h*K2/2, y[i] + h*L2/2],
L3 = g[t[i] + l/2, x[i] + l*K2/2, y[i] + l*L2/2],
K4 = f[t[i] + h, x[i] + h*K3, y[i] + h*L3],
L4 = g[t[i] + l, x[i] + l*K3, y[i] + l*L3],
x[i + 1] = x[i] + (K1 + 2 K2 + 2 K3 + K4)*h/6,
y[i + 1] = y[i] + (L1 + 2 L2 + 2 L3 + L4)*l/6},
{i,0,n}];
ListPlot[Table[{x[n], y[n]}, {n, 0, 10000}],
Frame -> True]
sol = Table[{x[n], y[n]}, {n, 1, 1000}];
p = Interpolation /@ Transpose@sol;
ParametricPlot[Evaluate@Through@p@t, {t, 1, 1000},
Frame -> True]
www.pdfgrip.com
0.5
0.0
–0.5
–1.0
–1.5
0.5
0.0
–0.5
–1.0
–1.5
Figures 6.13 and 6.14 show the RK4 solution (x(t), y(t)) of the modified
fractional Duffing system without and with iteration, respectively.
Remark Starting from the fractional RK procedures introduced in this section, the
reader can establish similar procedures for the other RK methods, established for
the integer order cases [1].
www.pdfgrip.com
Clear["‘*"]
f[t_, x_, y_, z_] := y;
g[t_, x_, y_, z_] := z;
p[t_, x_, y_, z_] := -z - x - 10^(-9) (Exp[y] - 1);
a = 0.998;
d = 0.01;
x[0] = 0;
y[0] = 0.4;
z[0] = 0;
tmax = 10000;
h = (d)^a/Gamma[a + 1];
Do[{k1 = h f[t[n], x[n], y[n], z[n]];
l1 = h g[n*d, x[n], y[n], z[n]];
m1 = h p[n*d, x[n], y[n], z[n]];
k2 = h f[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
l2 = h g[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
m2 = h p[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
50
0 20
–50 0
–50 –20
0
–40
20 –50
0
–20 0
–40
50
–50
Fig. 6.16 RK4 solution of the fractional Colpitts oscillator system with iteration
Solution
Clear["‘*"]
f[t_, x_, y_, z_] := y;
g[t_, x_, y_, z_] := z;
p[t_, x_, y_, z_] := -x - y - Sign[1 + 4*y];
a = 0.998; d = 0.01; x[0] = -0.5; y[0] = 0;
z[0] = 0; tmax = 2000; h = (d)^a/Gamma[a + 1];
Do[{k1 = h f[t[n], x[n], y[n], z[n]];
l1 = h g[n*d, x[n], y[n], z[n]];
m1 = h p[n*d, x[n], y[n], z[n]];
k2 = h f[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
l2 = h g[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
m2 = h p[n*d + h/2, x[n] + k1/2, y[n] + l1/2,
z[n] + m1/2];
www.pdfgrip.com
Figures 6.17 and 6.18 show the RK4 solution of the fractional Sprott oscillator
system without and with iteration, respectively.
0 0
5
www.pdfgrip.com
–2 –1
2 0 1
0
–2
–4
5
Fig. 6.18 RK4 solution of the Sprott oscillator system with iteration
for α = 0.998.
Solution In Mathematica the solution will be:
Clear["‘*"]
a = 0.998
RK4[f_, x0_, A_, B_, n_] :=
www.pdfgrip.com
Figure 6.19 shows the (x1 (t), x2 (t)) vector solution of the Van der Pol system.
–2 –1 1 2
–2
–4
www.pdfgrip.com
Example 2 Solve the fractional Rössler attractor system, using the vector RK4
method:
⎧
⎪
⎪ D 0.98 x1 (t) = −x2 − x3 , x1 (0) = 0,
⎪
⎪
⎪
⎨
D 0.98 x2 (t) = x1 + 0.2x2 , x2 (0) = 1,
⎪
⎪
⎪
⎪
⎪
⎩ 0.98
D x3 (t) = 0.2 + x3 (x1 − 8), x3 (0) = 0.
Solution
Clear["‘*"]
a = 0.998
RK4[f_, x0_, A_, B_, n_] :=
Module[{h, K1, K2, K3, K4, Sol = {{A, x0}},
x = x0, t}, h = (B - A)/n;
Do[t = A + k h^a/Gamma[a + 1];
K1 = h^a/Gamma[a + 1] f[t, x];
K2 = h^a/Gamma[a + 1] f[t +
(1/2) h^a/Gamma[a + 1], x + (1/2) K1];
K3 = h^a/Gamma[a + 1] f[t +
(1/2) h^a/Gamma[a + 1], x + (1/2) K2];
K4 = h^a/Gamma[a + 1] f[t +
h^a/Gamma[a + 1], x + K3];
x = x + (1/6) K1 + (1/3) K2 +
(1/3) K3 + (1/6) K4;
Sol = Append[Sol, {t, x}], {k, 1, n}];
Sol]
F[t_, x_] := {-x[[2]] - x[[3]], x[[1]]
+ 0.2 x[[2]], 0.2 + x[[1]] x[[3]] -
5.7 x[[3]]};
Figure 6.20 shows the (x1 (t), x2 (t)) vector solution of the Rössler attractor.
Example 3 (Volta Fractional Attractor) Find the solution of the fractional Volta
attractor using the vector RK4 algorithm:
⎧
⎪
⎪ D 0.998 x1 (t) = −x1 − 5x2 − x2 x3 , x1 (0) = 8,
⎪
⎪
⎪
⎨
D 0.998 x2 (t) = −85x1 − x2 − x1 x3 , x2 (0) = 2,
⎪
⎪
⎪
⎪
⎪
⎩ D 0.998 x (t) = 0.5x + x x + 1,
3 3 1 2 x3 (0) = 1.
www.pdfgrip.com
0.4
0.3
0.2 5
0.1
0.0 0
–5
0 –5
10 –10
Fig. 6.20 The (x1 (t), x2 (t), x3 (t)) vector solution of the Rössler attractor
Solution
In Mathematica can be written:
Clear["‘*"]
a = 0.998
RK4[f_, x0_, A_, B_, n_] :=
Module[{h, K1, K2, K3, K4, Sol = {{A, x0}},
x = x0, t}, h = (B - A)/n;
Do[t = A + k h^a/Gamma[a + 1];
K1 = h^a/Gamma[a + 1] f[t, x];
K2 = h^a/Gamma[a + 1] f[t + (1/2) h^a/Gamma[a + 1],
x + (1/2) K1];
K3 = h^a/Gamma[a + 1] f[t + (1/2) h^a/Gamma[a + 1],
x + (1/2) K2];
K4 = h^a/Gamma[a + 1] f[t + h^a/Gamma[a + 1],
x + K3];
x = x + (1/6) K1 + (1/3) K2 + (1/3) K3 + (1/6) K4;
Sol = Append[Sol, {t, x}], {k, 1, n}];
Sol]
www.pdfgrip.com
40
–50
20
–100
0
0
20 –20
40
–40
60
Fig. 6.21 The (x1 (t), x2 (t), x3 (t)) vector solution of the Volta attractor
www.pdfgrip.com
40
30
20 20
10
0
0
–10
10
–20
20
Fig. 6.22 The (x1 (t), x2 (t), x3 (t)) vector solution of the Lorenz attractor
Solution
Clear["‘*"]
a = 0.998
RK4[f_, x0_, A_, B_, n_] :=
Module[{h, K1, K2, K3, K4, Sol = {{A, x0}},
x = x0, t}, h = (B - A)/n;
Do[t = A + k h^a/Gamma[a + 1];
K1 = h^a/Gamma[a + 1] f[t, x];
K2 = h^a/Gamma[a + 1] f[t + (1/2) h^a/Gamma[a + 1],
x + (1/2) K1];
K3 = h^a/Gamma[a + 1] f[t + (1/2) h^a/Gamma[a + 1],
x + (1/2) K2];
K4 = h^a/Gamma[a + 1] f[t + h^a/Gamma[a + 1],
x + K3];
x = x + (1/6) K1 + (1/3) K2 + (1/3) K3 + (1/6) K4;
Sol = Append[Sol, {t, x}], {k, 1, n}];
Sol]
F[t_, x_] := {10 (x[[2]] - x[[1]]), x[[1]]
(28 - x[[3]])
www.pdfgrip.com
References 185
Figure 6.22 shows the (x1 (t), x2 (t)) vector solution of the Lorenz attractor.
References
Index
B F
Bernoulli numbers, 129 Fourth order Runge–Kutta method, 153
Bessel function of first kind, 115 Fractional calculus, v
Bessel generalized fractional differential Fractional Colpitts oscillator, 175
equation, 113 Fractional Cosine, 87
Beta function, 7 Fractional derivative, v
Fractional differential equation, vi, 47
Fractional Duffing system, 159
Fractional Rössler attractor system, 181
C
Fractional Rössler system, 163
Caputo, 24
Fractional Sine, 87
Caputo FDE, 47
Fractional Sprott oscillator, 176
Caputo fractional derivative, 24, 44
Fractional Van der Pol system, 157
Caputo fractional integral, 43
Fractional Volta attractor, 166, 181
Cauchy formula, 23
Chebyshev norm, 48
Chua attractor, 167 G
Complementary error function, 11 Galerkin method, 135
Cornu fractional differential system, 88 Gamma function, 1
Cornu Fractional System, 87 Gauss’ formula, 5
Gauss multiplication formula, 11
Generalized MacLaurin series, 109
D Ghelfand fractional differential equation, 73
Dirichlet equality, 21 Gram determinant, 96
Dirichlet theorem, 18
H
Hermite generalized fractional differential
E equation, 110
E function, 13
Error function, 11
Euler integral, 1 I
Euler method, 143 Imaginary error function, 11
Euler’s constant, 6 Inverse Laplace transform, 33
Exponential integral function, 11 Iterative Euler formula, 143
188 Index
L R
Lagrange multiplier, 122 Residues, 35
Lane and Emden fractional differential Riemann, 17
equation, 104 Riemann–Liouville derivative, 17
Laplace transform, 33 Riemann–Liouville FDE, 47
Least squares method, 124 Riemann–Liouville fractional derivative, 43
Left fractional integral, 17 Riemann–Liouville fractional integral, 43
Legendre duplication, 10 Right fractional integral, 17
Legendre fractional differential equation, 111 Runge–Kutta method, 150, 153
Liouville, 17
Lorenz attractor, 148, 162
Lorenz attractor system, 183 S
Lorenz fractional system, 118 Second order Runge–Kutta algorithm, 150
Lotka system, 116 Stirling formula, 37
T
M Taylor, B., 25
Method of Picard, 52 Taylor generalized formula, 110
Mittag-Leffler, 11 Two parameter Mittag-Leffler function, 11
Modified fractional Duffing system, 173
V
Van Der Pol fractional equation, 179
O Variational iteration method, 121
One parameter Mittag-Leffler function, 11 Vectorial Runge–Kutta algorithm, 179
Voltera integral, 48
P W
Post formula, 36 Weierstrass form, 6
Power series method, 94 Weierstrass test, 48, 51